# Nemotron

Nemotron is a family of open NVIDIA models with open weights, training data, and recipes for building specialized AI agents.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Nvidia expands AI infrastructure push into high-performance computing

DevFeed: [Nvidia expands AI infrastructure push into high-performance computing](<https://devfeed.tech/articles/nvidia-expands-ai-infrastructure-push-into-high-performance-computing-141282.md>)

Original publisher: [Read original article](<https://www.networkworld.com/article/4250303/nvidia-expands-ai-infrastructure-push-into-high-performance-computing.html>)

Author: Andy Patrizio

Published: 2026-11-18T17:10:21Z

Content type: news

Language: en

Sources: [Network World](<https://devfeed.tech/sources/network-world.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [Supercomputing](<https://devfeed.tech/topics/supercomputing.md>), [CUDA-X](<https://devfeed.tech/topics/cuda-x.md>), [systems](<https://devfeed.tech/topics/systems.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>), [AI Networking](<https://devfeed.tech/topics/ai-networking.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [ai-workloads](<https://devfeed.tech/tags/ai-workloads.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-cpus-and-processors-industry-markets-networking-technology-industry](<https://devfeed.tech/tags/artificial-intelligence-cpus-and-processors-industry-markets-networking-technology-industry.md>), [computing](<https://devfeed.tech/tags/computing.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cpus](<https://devfeed.tech/tags/cpus.md>), [cpus-and-processors](<https://devfeed.tech/tags/cpus-and-processors.md>), [cuda](<https://devfeed.tech/tags/cuda.md>), [data](<https://devfeed.tech/tags/data.md>), [data-processing](<https://devfeed.tech/tags/data-processing.md>), [edition-us](<https://devfeed.tech/tags/edition-us.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [high-performance](<https://devfeed.tech/tags/high-performance.md>), [high-performance-computing](<https://devfeed.tech/tags/high-performance-computing.md>), [industry](<https://devfeed.tech/tags/industry.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [markets](<https://devfeed.tech/tags/markets.md>), [networking](<https://devfeed.tech/tags/networking.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [performance](<https://devfeed.tech/tags/performance.md>), [storytype-news](<https://devfeed.tech/tags/storytype-news.md>), [technology-industry](<https://devfeed.tech/tags/technology-industry.md>)

### AI overview

Nvidia's SC26 announcements extend its push from GPU sales toward full HPC and AI infrastructure. Rubin HPC, expected in 2028, targets scientific simulations, engineering, and data processing with more than 50,000 CUDA cores, seventh-generation NVLink, and sparse-data optimizations. A rack-scale Vera CPU system combines 192 CPUs per rack with Quantum-X800 InfiniBand and SHARP, using NVL72 chassis, power, and cooling components. Nemotron for Science offers open models, datasets, evaluation tools, and post-training recipes for scientific AI agents that can interpret research, generate code, and use computational tools. CUDA-X updates include claimed speedups ranging from 3x to 900x for specific scientific tasks, with comparisons varying by library and baseline. Lambda is deploying Nvidia Photonics co-packaged optics switches to connect GB300 NVL72 racks containing over 10,000 GPUs; Nvidia presents lower power use, reliability, and scalability as intended benefits.

### Source excerpt

Perhaps anxious to talk about technology instead of its CEO drama, Nvidia unveiled a raft of HPC-related news at the Supercomputing '26 show taking place in Chicago. Nvidia's announcements at SC26 continue the company's efforts to expand beyond selling GPUs into supplying complete AI and high-performance computing infrastructure. Rubin HPC: A new GPU for scientific computing The news covers next-generation GPU and CPU platforms, networking and interconnect technology, accelerated scientific computing, and systems designed for large-scale AI workloads. Here is a breakdown of the major developments. Nvidia announced Rubin HPC, a new GPU within its Vera Rubin platform designed specifically for high-performance computing applications. Unlike GPUs primarily optimized for AI training and inference, Rubin HPC is designed to boost scientific simulations, engineering applications, and large-scale data processing. Rubin HPC has more than 50,000 CUDA cores on a single GPU. The core count for regular Roman GPUs has not been disclosed. It includes seventh-generation NVLink for accelerated GPU-to-GPU communication, sparse-data acceleration, and optimizations for sparse matrix operations, stencil computations, and sorting and reduction operations. Rubin HPC is expected to ship in 2028. Vera CPU platform expands into HPC Nvidia also announced a rack-scale version of its Vera CPU platform specifically designed for HPC. When it first introduced its own CPU, Nvidia positioned it as an alternative to traditional server processors from Intel and AMD. Now Nvidia is positioning its CPU for a specific market. The new configuration combines 192 Vera CPUs per rack plus Quantum-X800 InfiniBand networking, and SHARP in-network computing technology. It uses existing NVL72 rack components, including chassis, power distribution, and cooling infrastructure. Nemotron for Science: AI agents for scientific research Nemotron is a family of AI models developed by Nvidia, designed to power AI applicatio

## What RL on Expert-Built Tasks Teaches Nemotron 3.5 Super

DevFeed: [What RL on Expert-Built Tasks Teaches Nemotron 3.5 Super](<https://devfeed.tech/articles/what-rl-on-expert-built-tasks-teaches-nemotron-3-5-super-139598.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.22161>)

Author: Sushant Mehta, Edwin Chen

Published: 2026-11-18T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>), [arXiv: Software Engineering](<https://devfeed.tech/sources/arxiv-software-engineering.md>)

Topics: [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [post-training](<https://devfeed.tech/topics/post-training.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>)

Tags: [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-se](<https://devfeed.tech/tags/cs-se.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-3-5](<https://devfeed.tech/tags/nemotron-3-5.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [rl](<https://devfeed.tech/tags/rl.md>), [software-engineering](<https://devfeed.tech/tags/software-engineering.md>), [tool-use](<https://devfeed.tech/tags/tool-use.md>)

### AI overview

Reinforcement learning on five expert-built task datasets improved NVIDIA's Nemotron 3.5 Super across the reported benchmarks, with gains from 4.5 points on IFBench to 13.8 on GDP_pdf (10.4% to 24.2%). The study trained separate rank-32 LoRA adapters with CISPO on Tinker, without a supervised stage, and compared each adapter with the base model under the same task settings. Reported gains include SWE-bench Pro V2 (86.9% to 92.1%), Terminal-Bench 2.1 (63.9% to 70.3%), MultiChallenge (58.3% to 66.6%), AutomationBench (14.8% to 27.5%) and Chartography (39.4% to 50.6%). The document adapter also gained 8 points on AutomationBench despite training without API tools. On newly solved tasks, adapters more often addressed observed base-model failures: locating data and rules before tool use, creating test or scratch files in coding tasks, retaining earlier constraints, measuring chart marks in code, and reading spreadsheets.

### Source excerpt

arXiv:2611.22161v1 Announce Type: cross Abstract: We post-train NVIDIA's Nemotron 3.5 Super (120B total, 12B active parameters) with reinforcement learning (RL) on five expert-built task datasets, one at a time: real-world software engineering, complex constraints, long-horizon tool use, professional chart understanding and enterprise document reasoning. Each run trains a fresh rank-32 LoRA adapter with CISPO on Tinker, with no supervised stage, and each adapter is compared with the base model on the same tasks, harness and settings. On twelve headline benchmarks, eight built by other groups and four by Surge AI, the gains range from +4.5 points on IFBench to +13.8 on GDP_pdf, which rises from 10.4% to 24.2%; SWE-bench Pro V2 rises from 86.9% to 92.1%, Terminal-Bench 2.1 from 63.9% to 70.3%, MultiChallenge from 58.3% to 66.6%, AutomationBench from 14.8% to 27.5% and Chartography from 39.4% to 50.6%. The document adapter, trained without API tools, also gains +8.0 points on AutomationBench. In attempts read side by side with the base model's on tasks each adapter newly solves, the base model's failures recur in a few modes, and each adapter does more often what the base model missed: the tool-use adapter finds the data and the rules before acting, the coding adapter writes its own test or scratch files on SWE-bench Pro V2, the constraints adapter carries earlier turns into its answers, the chart adapter measures the marks in code, and the document adapter reads AutomationBench's spreadsheets.

## Nvidia says its open-source AI strategy remains intact after Jensen Huang's exit

DevFeed: [Nvidia says its open-source AI strategy remains intact after Jensen Huang's exit](<https://devfeed.tech/articles/nvidia-says-its-open-source-ai-strategy-remains-intact-after-jensen-huang-s-exit-138647.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/17/nvidia-says-its-open-source-ai-strategy-remains-intact-after-jensen-huangs-exit/>)

Author: Sarah Perez

Published: 2026-11-17T23:04:35Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [acquisition](<https://devfeed.tech/tags/acquisition.md>), [adoption](<https://devfeed.tech/tags/adoption.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-systems](<https://devfeed.tech/tags/agentic-ai-systems.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-adoption](<https://devfeed.tech/tags/ai-adoption.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [ai-technology](<https://devfeed.tech/tags/ai-technology.md>), [alice-huang](<https://devfeed.tech/tags/alice-huang.md>), [analysts](<https://devfeed.tech/tags/analysts.md>), [ceo](<https://devfeed.tech/tags/ceo.md>), [chinese-spy](<https://devfeed.tech/tags/chinese-spy.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [interim-ceo](<https://devfeed.tech/tags/interim-ceo.md>), [jensen-huang](<https://devfeed.tech/tags/jensen-huang.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [tc](<https://devfeed.tech/tags/tc.md>)

### AI overview

Nvidia interim CEO Colette Kress said the company's open-model strategy and planned $12.9 billion acquisition of Hugging Face remain on track after Jensen Huang's departure. She cited a 35-fold year-over-year rise in weekly token volume on OpenRouter, including 3.5-fold growth over three months, as evidence of accelerating adoption. Kress also described Nemotron as OpenRouter's most widely used U.S. open model. The acquisition was not addressed directly in the earnings call's Q&A, so the excerpt supports Nvidia's stated plans but provides no further transaction detail.

### Source excerpt

Nvidia's interim CEO reaffirmed its open-source AI strategy and plans to acquire Hugging Face, despite Jensen Huang's sudden departure.

## Ian Buck Kicks Off NVIDIA's Week at SC26 as Science Enters the Age of Agents

DevFeed: [Ian Buck Kicks Off NVIDIA's Week at SC26 as Science Enters the Age of Agents](<https://devfeed.tech/articles/ian-buck-kicks-off-nvidia-s-week-at-sc26-as-science-enters-the-age-of-agents-138205.md>)

Original publisher: [Read original article](<https://blogs.nvidia.com/blog/ian-buck-sc26/>)

Author: Brian Caulfield

Published: 2026-11-17T14:00:10Z

Content type: news

Language: en

Sources: [NVIDIA](<https://devfeed.tech/sources/nvidia-blog.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [DGX Spark](<https://devfeed.tech/topics/dgx-spark.md>), [NVIDIA Vera](<https://devfeed.tech/topics/nvidia-vera.md>), [CUDA](<https://devfeed.tech/topics/cuda.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [code](<https://devfeed.tech/tags/code.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [events](<https://devfeed.tech/tags/events.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [high-performance](<https://devfeed.tech/tags/high-performance.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [ian-buck](<https://devfeed.tech/tags/ian-buck.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [rubin](<https://devfeed.tech/tags/rubin.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [science](<https://devfeed.tech/tags/science.md>), [supercomputing](<https://devfeed.tech/tags/supercomputing.md>)

### AI overview

NVIDIA's SC26 announcements combine locally runnable science agents with new HPC hardware and quantum-computing work. Nemotron for Science pairs the Nemotron 3.5 Super VL open model with an open chemistry dataset and post-training recipes; researchers can run it on DGX Spark, a workstation or their own supercomputer. Buck cited an Argonne agent that planned and tested hypotheses, selected simulations and identified two binding sites, plus examples for GROMACS repair, telescope data plotting and GPU porting. The Nsight AI Porting Agent reportedly sped up a weather-model routine by more than 5x. NVIDIA also announced Rubin HPC, supporting FP64 to FP4 and expected in 2028, and Vera CPUs selected for a German climate supercomputer expected in early 2028. For quantum research, NVIDIA, IBM and Fermilab plan a co-design platform; Buck cited a CUDA-Q decoding algorithm that reduced a research computation from 12 weeks to 45 minutes. These are reported announcements and results from the excerpt.

### Source excerpt

NVIDIA's Ian Buck kicked off the company's week at SC26 in Chicago on Monday with new tools for AI agents, software that can plan experiments, run code and check results on researchers' own systems. In NVIDIA's special address, Buck announced Nemotron for Science, the Rubin HPC GPU and an expanded Vera CPU platform for high-performance [...]

## High-Throughput Serving of Nemotron 3.5 Transcribe with Regime-Aware Megakernels and Kernel Fusion

DevFeed: [High-Throughput Serving of Nemotron 3.5 Transcribe with Regime-Aware Megakernels and Kernel Fusion](<https://devfeed.tech/articles/high-throughput-serving-of-nemotron-3-5-transcribe-with-regime-aware-megakernels-and-kernel-fusion-136987.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20095>)

Author: Aditya Chatterjee, Sanjay Singh Chauhan, Mayank Jain, Arushi Dhingra, Sahin Chowdhury, Rahul Mittal

Published: 2026-11-17T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Distributed Computing](<https://devfeed.tech/sources/arxiv-distributed-computing.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [asr](<https://devfeed.tech/tags/asr.md>), [batching](<https://devfeed.tech/tags/batching.md>), [beats](<https://devfeed.tech/tags/beats.md>), [best](<https://devfeed.tech/tags/best.md>), [budget](<https://devfeed.tech/tags/budget.md>), [can](<https://devfeed.tech/tags/can.md>), [changes](<https://devfeed.tech/tags/changes.md>), [completion](<https://devfeed.tech/tags/completion.md>), [concurrency](<https://devfeed.tech/tags/concurrency.md>), [cs-dc](<https://devfeed.tech/tags/cs-dc.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-pf](<https://devfeed.tech/tags/cs-pf.md>), [eess-as](<https://devfeed.tech/tags/eess-as.md>), [fusion](<https://devfeed.tech/tags/fusion.md>), [kernel](<https://devfeed.tech/tags/kernel.md>), [megakernels](<https://devfeed.tech/tags/megakernels.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-3-5](<https://devfeed.tech/tags/nemotron-3-5.md>), [nemotron-3-5-transcribe](<https://devfeed.tech/tags/nemotron-3-5-transcribe.md>), [serving](<https://devfeed.tech/tags/serving.md>), [throughput](<https://devfeed.tech/tags/throughput.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

The serving engine adapts kernel fusion per inference pass, choosing between persistent megakernels and fused-kernel chains using offline-measured crossover points as continuous batching changes the workload. On H100 SXM it reaches 4,345x real-time throughput (1.39x tuned vLLM at peak, and up to 4.39x faster at low concurrency); on B200 it reaches 6,367x (1.64x vLLM at the highest measured concurrency). WER is 1.91-1.93% versus vLLM's 1.88-1.93%. Within a 400 ms p95 completion-time budget, it serves eight times as many concurrent requests. Per-pass selection outperforms always using megakernels by 23% at 128 concurrent requests and always using fused kernels by 9% at eight. Opt-in options, including greedy MoE layer skipping, add 13-16% throughput on B200 for offline jobs that accept higher WER.

### Source excerpt

arXiv:2611.20095v1 Announce Type: cross Abstract: Nemotron 3.5 Transcribe ranks first among open-weight models on the Hugging Face Open ASR leaderboards. Serving it efficiently takes more than fast operators: continuous batching changes the amount and mix of work at every iteration, and the best fusion granularity changes with it. We present a serving engine that, for every pass, chooses between persistent megakernels and chains of fused kernels at crossovers measured offline. On an H100 SXM GPU, throughput of our accuracy-preserving serving peaks at 4,345x real time, 1.39x the peak of tuned vLLM, and is up to 4.39x faster at low concurrency; on B200 it peaks at 6,367x, 1.64x vLLM at the highest concurrency we measured. Word error rate (WER) is 1.91 to 1.93% against 1.88 to 1.93% for vLLM. Within a 400 ms budget for p95 end-to-end completion time, it serves 8x as many concurrent requests. Choosing per pass beats either level alone: 23% higher throughput than always using megakernels at 128 concurrent requests, and 9% higher than always using fused kernels at 8. Opt-in switches, including greedy MoE layer skipping, add 13 to 16% throughput on B200 for offline jobs that can accept higher WER.

## Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents

DevFeed: [Nemotron 3.5 Bolt E4B 4GB: An Open and Efficient Small Language Model for Fast Conversational Agents](<https://devfeed.tech/articles/nemotron-3-5-bolt-e4b-4gb-an-open-and-efficient-small-language-model-for-fast-conversational-agents-137611.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.20771>)

Author: NVIDIA, :, Adi Margolin, Aditya Malte, Aishwarya Chhabra, Alex Hodson, Ameya Sunil Mahabaleshwarkar, Ambrish Dantrey, Amy Shen, Andrii Skliar, Annamalai Chockalingam, Anusha Pant, Ashwath Aithal, Ashw

Published: 2026-11-17T05:00:00Z

Content type: release

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Conversational AI](<https://devfeed.tech/topics/conversational-ai.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agents](<https://devfeed.tech/tags/agents.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [capabilities](<https://devfeed.tech/tags/capabilities.md>), [concurrency](<https://devfeed.tech/tags/concurrency.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decode](<https://devfeed.tech/tags/decode.md>), [dflash](<https://devfeed.tech/tags/dflash.md>), [dgx-spark](<https://devfeed.tech/tags/dgx-spark.md>), [edge](<https://devfeed.tech/tags/edge.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [fast](<https://devfeed.tech/tags/fast.md>), [geforce](<https://devfeed.tech/tags/geforce.md>), [geforce-rtx](<https://devfeed.tech/tags/geforce-rtx.md>), [language-model](<https://devfeed.tech/tags/language-model.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-3-5](<https://devfeed.tech/tags/nemotron-3-5.md>), [nemotron-3-5-bolt](<https://devfeed.tech/tags/nemotron-3-5-bolt.md>), [open](<https://devfeed.tech/tags/open.md>), [small](<https://devfeed.tech/tags/small.md>), [small-language-model](<https://devfeed.tech/tags/small-language-model.md>)

### AI overview

Nemotron 3.5 Bolt E4B 4GB is an open small language model post-trained from Gemma-4-E4B-it for fast conversational agents, with reported improvements in multi-turn tool use, instruction following, and input grounding. Across 21 benchmarks, the authors report the highest average accuracy and fewest output tokens in reasoning and non-reasoning modes versus Gemma-4-E4B-it and Qwen3.5-4B. The release includes a DFlash drafter: at concurrency 1, it reaches 2.74x the per-user decode speed of inference without speculative decoding; at batch size 1, it achieves 1.21x Gemma-4-E4B-it's generation throughput with MTP. These figures reflect the configurations stated in the abstract.

### Source excerpt

arXiv:2611.20771v1 Announce Type: cross Abstract: We introduce Nemotron 3.5 Bolt E4B 4GB, an open and efficient small language model further post-trained from Gemma-4-E4B-it and optimized for speed and intelligence per GB of memory used for fast conversational agents. The model carries forward Gemma's strong conversational capabilities and further improves multi-turn tool use, instruction following, and input grounding for fast, local voice/device assistants, interactive personas, and embedded/Edge interfaces. The model combines natural conversation with multi-turn agentic capabilities to power voice/device assistants, interactive characters and robotic/IoT interfaces running on GeForce RTX, Jetson, and RTX/DGX Spark. Experiments on 21 benchmarks show that Nemotron 3.5 Bolt E4B 4GB achieves the highest average accuracy while generating the fewest output tokens in both reasoning and non-reasoning modes, compared with Gemma-4-E4B-it and Qwen3.5-4B. The model is released with a DFlash drafter. With DFlash, Bolt E4B achieves $2.74\times$ the per-user decode speed of inference without speculative decoding at concurrency~1, and $1.21\times$ the generation throughput of Gemma-4-E4B-it with MTP at batch size 1.

## At SC26, Nvidia debuts new Rubin HPC chip, scientific research libraries

DevFeed: [At SC26, Nvidia debuts new Rubin HPC chip, scientific research libraries](<https://devfeed.tech/articles/at-sc26-nvidia-debuts-new-rubin-hpc-chip-scientific-research-libraries-136387.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/16/nvidia-debuts-new-rubin-hpc-chip-scientific-research-libraries/>)

Author: Maria Deutscher

Published: 2026-11-17T01:45:14Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [AI for science](<https://devfeed.tech/topics/ai-for-science.md>), [CUDA-X](<https://devfeed.tech/topics/cuda-x.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [astrophysics](<https://devfeed.tech/tags/astrophysics.md>), [biology](<https://devfeed.tech/tags/biology.md>), [cuda-x](<https://devfeed.tech/tags/cuda-x.md>), [cuphoton](<https://devfeed.tech/tags/cuphoton.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [libraries](<https://devfeed.tech/tags/libraries.md>), [llm](<https://devfeed.tech/tags/llm.md>), [nasdaq-nvda](<https://devfeed.tech/tags/nasdaq-nvda.md>), [nemotron-3-5-super-vl](<https://devfeed.tech/tags/nemotron-3-5-super-vl.md>), [news](<https://devfeed.tech/tags/news.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvlink](<https://devfeed.tech/tags/nvlink.md>), [rubin](<https://devfeed.tech/tags/rubin.md>), [rubin-hpc](<https://devfeed.tech/tags/rubin-hpc.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [scientific-research](<https://devfeed.tech/tags/scientific-research.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-5](<https://devfeed.tech/tags/top-story-5.md>), [top-story-6](<https://devfeed.tech/tags/top-story-6.md>)

### AI overview

Nvidia announced a science-focused hardware and software lineup at SC26. Its planned Rubin HPC GPU, due in 2028, will have more than 50,000 CUDA cores, a new NVLink iteration, and sparse-data acceleration; the current Rubin GPU has over 28,000 CUDA cores. New CUDA-X tools include cuPhoton.xBeam for Event Horizon Telescope analysis, which Nvidia says can reduce tasks from a month to eight minutes, cuPhoton.xLens for fusion-plasma simulation, and open-source cuSpa for cell-polygon calculations. Nvidia also packaged its Nemotron 3.5 Super VL model with scientific post-training workflows and nearly 50,000 chemistry questions. CUDA-Q additions include quantum software components and an error-mitigation algorithm; Nvidia, IBM, and Fermilab are also developing a platform combining DGX B200 systems with IBM quantum hardware.

### Source excerpt

Nvidia Corp. today introduced a set of offerings designed to speed up scientific research across a range of fields. The lineup is headlined by Rubin HPC, an upcoming graphics card with significantly more cores than the company's current flagship chip. It's joined by more than a half-dozen software tools geared toward scientists and engineers. The [...] The post At SC26, Nvidia debuts new Rubin HPC chip, scientific research libraries appeared first on SiliconANGLE.

## Post-Training NVIDIA Nemotron to Increase Accuracy in Your Scientific Domain

DevFeed: [Post-Training NVIDIA Nemotron to Increase Accuracy in Your Scientific Domain](<https://devfeed.tech/articles/post-training-nvidia-nemotron-to-increase-accuracy-in-your-scientific-domain-136679.md>)

Original publisher: [Read original article](<https://developer.nvidia.com/blog/post-training-nvidia-nemotron-to-increase-accuracy-in-your-scientific-domain/>)

Author: Gautham Dharuman

Published: 2026-11-16T16:00:00Z

Content type: tutorial

Language: en

Sources: [NVIDIA Developer](<https://devfeed.tech/sources/nvidia-developer.md>), [NVIDIA Technical Blog](<https://devfeed.tech/sources/nvidia-technical-blog.md>)

Topics: [post-training](<https://devfeed.tech/topics/post-training.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [rlvr](<https://devfeed.tech/topics/rlvr.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [AI for science](<https://devfeed.tech/topics/ai-for-science.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [NeMo](<https://devfeed.tech/topics/nemo.md>)

Tags: [17](<https://devfeed.tech/tags/17.md>), [adapter](<https://devfeed.tech/tags/adapter.md>), [adapters](<https://devfeed.tech/tags/adapters.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-generative-ai](<https://devfeed.tech/tags/agentic-ai-generative-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-agents](<https://devfeed.tech/tags/autonomous-agents.md>), [computational-chemistry-materials-science](<https://devfeed.tech/tags/computational-chemistry-materials-science.md>), [developer-tools-techniques](<https://devfeed.tech/tags/developer-tools-techniques.md>), [featured](<https://devfeed.tech/tags/featured.md>), [innovation](<https://devfeed.tech/tags/innovation.md>), [machine-learning-artificial-intelligence](<https://devfeed.tech/tags/machine-learning-artificial-intelligence.md>), [natural-language](<https://devfeed.tech/tags/natural-language.md>), [nemo](<https://devfeed.tech/tags/nemo.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-nemotron](<https://devfeed.tech/tags/nvidia-nemotron.md>), [post-training](<https://devfeed.tech/tags/post-training.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [research](<https://devfeed.tech/tags/research.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [scientific-computing](<https://devfeed.tech/tags/scientific-computing.md>), [scientific-research](<https://devfeed.tech/tags/scientific-research.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

The excerpt describes adapting Nemotron 3.5 Super VL to scientific tasks with reinforcement learning from verifiable rewards. Experimental records become prompts paired with reference outcomes; a task-specific verifier scores model submissions, so the reward comes from data without human preference labels. In the drug-metabolism example, Morgan fingerprints and Tanimoto similarity score predicted metabolites. The recipe uses NeMo Gym for the environment and NeMo RL for GRPO training with LoRA. It reports SFT improved accuracy by 12%, with RL adding 17% at a 16,384-token output budget. An optional warm start distills teacher traces that meet a similarity threshold; the excerpt reports retaining 1,123 of 1,934 prompts at a threshold of at least 0.7. It emphasizes choosing a reward that reflects the scientific outcome and evaluating on held-out data. These figures and implementation details are specific to the described example.

### Source excerpt

Agentic AI is reshaping scientific research and innovation. Researchers communicate complex scientific hypotheses in natural language and let autonomous agents...

## Route coding agent requests to the right model on Amazon Bedrock with NVIDIA NeMo Switchyard

DevFeed: [Route coding agent requests to the right model on Amazon Bedrock with NVIDIA NeMo Switchyard](<https://devfeed.tech/articles/route-coding-agent-requests-to-the-right-model-on-amazon-bedrock-with-nvidia-nemo-switchyard-133426.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/machine-learning/route-coding-agent-requests-to-the-right-model-on-amazon-bedrock-with-nvidia-nemo-switchyard/>)

Author: Aswath Ram A Srinivasan

Published: 2026-11-13T22:30:12Z

Content type: tutorial

Language: en

Sources: [Artificial Intelligence](<https://devfeed.tech/sources/artificial-intelligence.md>)

Topics: [litellm](<https://devfeed.tech/topics/litellm.md>), [NeMo](<https://devfeed.tech/topics/nemo.md>), [agentic-coding](<https://devfeed.tech/topics/agentic-coding.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Claude Code](<https://devfeed.tech/topics/claude-code.md>), [codex](<https://devfeed.tech/topics/codex.md>)

Tags: [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [amazon-bedrock](<https://devfeed.tech/tags/amazon-bedrock.md>), [aws](<https://devfeed.tech/tags/aws.md>), [aws-cdk](<https://devfeed.tech/tags/aws-cdk.md>), [coding](<https://devfeed.tech/tags/coding.md>), [coding-agent](<https://devfeed.tech/tags/coding-agent.md>), [coding-agents](<https://devfeed.tech/tags/coding-agents.md>), [cost](<https://devfeed.tech/tags/cost.md>), [litellm](<https://devfeed.tech/tags/litellm.md>), [nemo](<https://devfeed.tech/tags/nemo.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [route](<https://devfeed.tech/tags/route.md>), [routing](<https://devfeed.tech/tags/routing.md>), [solved](<https://devfeed.tech/tags/solved.md>), [switchyard](<https://devfeed.tech/tags/switchyard.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [technical-how-to](<https://devfeed.tech/tags/technical-how-to.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

The article's main takeaway is to judge coding-agent routing by cost per solved task, since multi-turn history, retries, and failed runs make token price alone misleading. It walks through deploying LiteLLM with NVIDIA NeMo Switchyard on Amazon Bedrock using AWS CDK, then comparing a routed model group with single-model baselines. Stage routing uses recent tool results to choose between exactly two candidates; it starts with the efficient model when signals are inconclusive and can escalate after failures. Random routing splits requests among candidates by weight. The sample's benchmark estimates cost from token usage and a price table, and counts a task solved when its tests pass within the turn limit. Results are illustrative: repeat runs, inspect hard tasks and truncation, and treat costs as estimates rather than billing data. The Switchyard integration is experimental, supports only specified message shapes and Chat Completions paths, and is presented for non-production use.

### Source excerpt

Coding agents work in tasks, and the number that matters is cost per solved task, not price per token. In this post, you use the AWS CDK to deploy LiteLLM with the NVIDIA NeMo Switchyard routing plugin on Amazon Bedrock, then measure cost per solved task for a routed model group against single-model baselines.

## Jensen Huang's exit puts Nvidia's support for open source in question

DevFeed: [Jensen Huang's exit puts Nvidia's support for open source in question](<https://devfeed.tech/articles/jensen-huang-s-exit-puts-nvidia-s-support-for-open-source-in-question-130764.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/12/jensen-huangs-exit-puts-nvidias-support-for-open-source-in-question/>)

Author: Rebecca Bellan

Published: 2026-11-12T18:39:23Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Open Source](<https://devfeed.tech/topics/open-source.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-systems](<https://devfeed.tech/tags/agentic-ai-systems.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-coding](<https://devfeed.tech/tags/ai-coding.md>), [ai-companies](<https://devfeed.tech/tags/ai-companies.md>), [ai-developers](<https://devfeed.tech/tags/ai-developers.md>), [ai-labs](<https://devfeed.tech/tags/ai-labs.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [alice-huang](<https://devfeed.tech/tags/alice-huang.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [business-strategy](<https://devfeed.tech/tags/business-strategy.md>), [hugging-face](<https://devfeed.tech/tags/hugging-face.md>), [jensen-huang](<https://devfeed.tech/tags/jensen-huang.md>), [leadership](<https://devfeed.tech/tags/leadership.md>), [models](<https://devfeed.tech/tags/models.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [open-ai](<https://devfeed.tech/tags/open-ai.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>)

### AI overview

Jensen Huang's departure leaves Nvidia's support for open AI models uncertain, though the article describes commercial reasons the company may continue it. Huang backed Nvidia's Nemotron open-weight models, investments in open-model startups, and public opposition to tighter restrictions; the company also has a pending $13 billion Hugging Face acquisition and a $6 billion non-exclusive license for Poolside technology. As major customers develop their own chips, Nvidia has an incentive to attract more developers and organizations that train, fine-tune, and run models on its GPUs. CFO Colette Kress is interim CEO, but her approach to the Hugging Face deal and open models is unknown. The article also connects Huang's advocacy to debate over Chinese AI models and U.S. chip export controls, noting that restrictions could limit competition for American companies.

### Source excerpt

Jensen Huang's resignation raises questions about Nvidia's commitment to open source AI, including its Hugging Face acquisition, open-weight models, and advocacy against restrictions, as the company navigates a leadership transition.

## Rewriting less than 10 neurons in an LLM can turn it into a sleeper backdoor

DevFeed: [Rewriting less than 10 neurons in an LLM can turn it into a sleeper backdoor](<https://devfeed.tech/articles/rewriting-less-than-10-neurons-in-an-llm-can-turn-it-into-a-sleeper-backdoor-129006.md>)

Original publisher: [Read original article](<https://www.infoworld.com/article/4247378/rewriting-less-than-10-neurons-in-an-llm-can-turn-it-into-a-sleeper-backdoor-2.html>)

Author: Gyana Swain

Published: 2026-11-11T12:31:23Z

Content type: news

Language: en

Sources: [InfoWorld](<https://devfeed.tech/sources/infoworld.md>)

Topics: [ai security](<https://devfeed.tech/topics/ai-security.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [ML model versioning](<https://devfeed.tech/topics/ml-model-versioning.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [gemma](<https://devfeed.tech/topics/gemma.md>), [qwen](<https://devfeed.tech/topics/qwen.md>), [prompt injection](<https://devfeed.tech/topics/prompt-injection.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-model](<https://devfeed.tech/tags/ai-model.md>), [backdoor](<https://devfeed.tech/tags/backdoor.md>), [development-tools](<https://devfeed.tech/tags/development-tools.md>), [development-tools-security-software-development](<https://devfeed.tech/tags/development-tools-security-software-development.md>), [gemma-4-31b](<https://devfeed.tech/tags/gemma-4-31b.md>), [integrity](<https://devfeed.tech/tags/integrity.md>), [language-en](<https://devfeed.tech/tags/language-en.md>), [lazarus](<https://devfeed.tech/tags/lazarus.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-research](<https://devfeed.tech/tags/microsoft-research.md>), [model](<https://devfeed.tech/tags/model.md>), [model-weights](<https://devfeed.tech/tags/model-weights.md>), [models](<https://devfeed.tech/tags/models.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [open-weight-ai](<https://devfeed.tech/tags/open-weight-ai.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [payload](<https://devfeed.tech/tags/payload.md>), [permissions](<https://devfeed.tech/tags/permissions.md>), [qwen](<https://devfeed.tech/tags/qwen.md>), [security](<https://devfeed.tech/tags/security.md>), [software-development](<https://devfeed.tech/tags/software-development.md>), [storytype-news](<https://devfeed.tech/tags/storytype-news.md>), [trigger](<https://devfeed.tech/tags/trigger.md>)

### AI overview

Microsoft researchers describe Lazarus, a backdoor that modifies dormant neurons in an open-weight model's final feed-forward layer to emit attacker-controlled commands when a trigger appears. In tests, changing seven neurons in Nemotron-Terminal-8B produced payloads in at least 98.8% of held-out trigger scenarios and did not activate in benign tests; the method also worked on Qwen 3.6-27B and Gemma-4-31B-IT. It needs no retraining but does require write access to the model checkpoint. Backdoored models may pass standard evaluations. Agent-built scanners flagged 91.4% of held-out altered checkpoints with a 9.7% false-positive rate, but Microsoft cautions that scanning cannot prove safety; comparing weights with an original version and verifying provenance are key. Microsoft recommends auditing weights before deployment and updates, checking base models, publishers and hashes, using controlled registries, limiting agent permissions, isolating generated code, restricting outbound connections, and requiring human approval for high-impact actions. Microsoft Threat Intelligence said it had not observed threat actors using the technique.

### Source excerpt

Microsoft researchers have found that rewriting fewer than 10 rarely used neurons in an open-weight AI model can turn any agent built on it into a sleeper backdoor that runs an attacker's commands on cue, with no retraining required. Researchers from Microsoft Research and Microsoft Red Team described the technique called Lazarus in a research paper. It modifies dormant neurons in a model's final feed-forward layer to recognize a trigger and inject attacker-controlled commands into the model's output. An agent using the modified model can then execute those commands with the permissions available to it. "The trigger can be as mundane as a date change, e.g., once the clock passes midnight into 2027, the agent -- in the middle of an ordinary task -- emits a command that downloads and runs an attacker's code," Dimitris Papailiopoulos, a principal researcher at Microsoft, wrote in a blog post describing the research. "The agent then executes it with whatever access you gave it." Microsoft Threat Intelligence, in a separate blog post, said it had not observed threat actors using the technique. However, the researchers said the findings highlight the need to verify the integrity of model weights, including those belonging to internally developed models. Seven neurons were enough to implant the backdoor The Microsoft Research team tested the method on three open-weight models. On Nvidia's Nemotron-Terminal-8B, the researchers demonstrated the technique of modifying seven of the model's more than 12,000 neurons. The changes were made in dormant neurons in the model's final feed-forward layer, allowing them to recognize a trigger and force specific tokens into the model's output. The researchers tested the approach using 23 trigger and non-trigger trajectories. The modified model generated the malicious payload in at least 98.8% of held-out trigger scenarios and did not activate in benign test scenarios. The technique also worked against Qwen 3.6-27B and Gemma-4-31B-IT, sugges

## Hibana: Fast Agentic Inference on Unified-Memory Devices

DevFeed: [Hibana: Fast Agentic Inference on Unified-Memory Devices](<https://devfeed.tech/articles/hibana-fast-agentic-inference-on-unified-memory-devices-125952.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.10469>)

Author: Yahya Emara, Chi-Chih Chang, Anthony Fei, Ahmed F. AbouElhamayed, Chong-Yan Chen, Hung-Yueh Chiang, Kai-Chiang Wu, Mohamed S. Abdelfattah

Published: 2026-11-10T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Local AI](<https://devfeed.tech/topics/local-ai.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [vllm](<https://devfeed.tech/topics/vllm.md>), [DGX Spark](<https://devfeed.tech/topics/dgx-spark.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [llama.cpp](<https://devfeed.tech/topics/llama-cpp.md>)

Tags: [adaptive](<https://devfeed.tech/tags/adaptive.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-tasks](<https://devfeed.tech/tags/agentic-tasks.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [code](<https://devfeed.tech/tags/code.md>), [coding](<https://devfeed.tech/tags/coding.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [data-movement](<https://devfeed.tech/tags/data-movement.md>), [datacenter](<https://devfeed.tech/tags/datacenter.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [devices](<https://devfeed.tech/tags/devices.md>), [dgx](<https://devfeed.tech/tags/dgx.md>), [fast](<https://devfeed.tech/tags/fast.md>), [inference](<https://devfeed.tech/tags/inference.md>), [memory](<https://devfeed.tech/tags/memory.md>), [unified-memory](<https://devfeed.tech/tags/unified-memory.md>)

### AI overview

Hibana targets local agent workloads on unified-memory devices by co-optimizing model representation, memory management, and decoding. Its model surgery reduces memory footprint and data movement; unified-memory management preserves agent state across turns for quick resumption; and adaptive speculation generates multiple tokens per step. The design accounts for shared memory capacity and limited bandwidth. On one DGX Spark, serving Qwen and Nemotron models up to 125B parameters, Hibana reports up to 3.4x more agentic tasks per hour and up to 3.9x higher decoding throughput than established frameworks including vLLM, NVIDIA Flare, llama.cpp, and TensorFold. These are reported maxima for the tested setup.

### Source excerpt

arXiv:2611.10469v1 Announce Type: cross Abstract: Local AI promises capable personal assistants that run on users' own devices, keeping data and computation under their control. Open-weight models that fit on local devices can now handle practical tasks such as coding. Yet current inference engines fall short. Popular engines such as vLLM and SGLang target datacenter GPUs with dedicated host memory, while local engines underuse the memory hierarchy of unified-memory devices. We present Hibana, an inference engine for unified-memory devices that co-optimizes model representation, memory management, and decoding strategy. Model surgery, the process of editing a trained model's architecture to suit its hardware, changes how the model computes, shrinking its memory footprint and data movement. Unified-memory management preserves each agent's state across turns, so long-running agents can resume instantly instead of recomputing their state. Adaptive speculation generates several tokens per step. Hibana codesigns these features around the shared memory capacity and limited bandwidth of unified-memory systems-on-chip to support local multi-agent workloads. On a single DGX Spark serving Qwen and Nemotron models of up to 125B parameters, Hibana completes up to 3.4$\times$ more agentic tasks per hour and delivers up to 3.9$\times$ higher decoding throughput than other established frameworks such as vLLM, NVIDIA Flare, llama.cpp, and TensorFold. Code is available at https://github.com/sharar-project/hibana.

## NVIDIA Nemotron 3.5 Super VL is now available in Amazon SageMaker JumpStart

DevFeed: [NVIDIA Nemotron 3.5 Super VL is now available in Amazon SageMaker JumpStart](<https://devfeed.tech/articles/nvidia-nemotron-3-5-super-vl-is-now-available-in-amazon-sagemaker-jumpstart-125097.md>)

Original publisher: [Read original article](<https://aws.amazon.com/blogs/machine-learning/nvidia-nemotron-3-5-super-vl-is-now-available-in-amazon-sagemaker-jumpstart/>)

Author: Venu Kanamatareddy

Published: 2026-11-09T20:01:20Z

Content type: release

Language: en

Sources: [Artificial Intelligence](<https://devfeed.tech/sources/artificial-intelligence.md>)

Topics: [Amazon SageMaker JumpStart](<https://devfeed.tech/topics/amazon-sagemaker-jumpstart.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Amazon SageMaker](<https://devfeed.tech/topics/amazon-sagemaker.md>), [vlm](<https://devfeed.tech/topics/vlm.md>), [NVFP4](<https://devfeed.tech/topics/nvfp4.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [post-training](<https://devfeed.tech/topics/post-training.md>)

Tags: [advanced-300](<https://devfeed.tech/tags/advanced-300.md>), [agentic-workloads](<https://devfeed.tech/tags/agentic-workloads.md>), [agents](<https://devfeed.tech/tags/agents.md>), [amazon-sagemaker](<https://devfeed.tech/tags/amazon-sagemaker.md>), [amazon-sagemaker-jumpstart](<https://devfeed.tech/tags/amazon-sagemaker-jumpstart.md>), [announcements](<https://devfeed.tech/tags/announcements.md>), [bf16](<https://devfeed.tech/tags/bf16.md>), [computer-use](<https://devfeed.tech/tags/computer-use.md>), [document-intelligence](<https://devfeed.tech/tags/document-intelligence.md>), [image](<https://devfeed.tech/tags/image.md>), [inference](<https://devfeed.tech/tags/inference.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-3-5](<https://devfeed.tech/tags/nemotron-3-5.md>), [nvfp4](<https://devfeed.tech/tags/nvfp4.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-nemotron-3-5-super-vl](<https://devfeed.tech/tags/nvidia-nemotron-3-5-super-vl.md>), [text](<https://devfeed.tech/tags/text.md>), [variant](<https://devfeed.tech/tags/variant.md>), [video](<https://devfeed.tech/tags/video.md>), [vision](<https://devfeed.tech/tags/vision.md>), [vision-language-model](<https://devfeed.tech/tags/vision-language-model.md>)

### AI overview

NVIDIA Nemotron 3.5 Super VL is available through SageMaker JumpStart in NVFP4 and BF16 variants. It accepts text, images, and video and generates text; the 120B-parameter model has 12B active parameters and supports configurable reasoning. NVFP4 is the default option for G7 instances, while BF16 reference weights run on P5en with H200 GPUs and can be downloaded for customization. The reported performance measurements use different GPU systems, so they do not isolate the effect of precision. Deployment requires endpoint permissions and instance quota; new accounts start with zero G7 quota, and the SDK path requires SageMaker Python SDK v3.25 or later. For inference, media must be sent inline as base64 because endpoints have network isolation. PDFs must be rendered to images, and requests accept one image and one video. Context limits vary by serving configuration; video frame sampling is fixed at endpoint launch, and reasoning-enabled video requests need at least 8,192 max tokens. The deployed configuration includes an MTP head but does not enable speculative decoding.

### Source excerpt

NVIDIA Nemotron 3.5 Super VL, a 120B-parameter multimodal vision-language model, is now available in Amazon SageMaker JumpStart. Deploy the NVFP4 or BF16 variant in a few steps, then run text, image, and video inference for document intelligence, computer-use agents, and multimodal agentic workloads.

## Korea's AI Ecosystem Advances and Secures Agentic AI With NVIDIA

DevFeed: [Korea's AI Ecosystem Advances and Secures Agentic AI With NVIDIA](<https://devfeed.tech/articles/korea-s-ai-ecosystem-advances-and-secures-agentic-ai-with-nvidia-123630.md>)

Original publisher: [Read original article](<https://blogs.nvidia.com/blog/korea-secure-agentic-ai/>)

Author: Nicola Sessions

Published: 2026-11-09T05:00:39Z

Content type: news

Language: en

Sources: [NVIDIA](<https://devfeed.tech/sources/nvidia-blog.md>)

Topics: [OpenShell](<https://devfeed.tech/topics/openshell.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [NemoClaw](<https://devfeed.tech/topics/nemoclaw.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [blueprints](<https://devfeed.tech/tags/blueprints.md>), [employee](<https://devfeed.tech/tags/employee.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [hyundai](<https://devfeed.tech/tags/hyundai.md>), [intelligence](<https://devfeed.tech/tags/intelligence.md>), [korea](<https://devfeed.tech/tags/korea.md>), [models](<https://devfeed.tech/tags/models.md>), [nemoclaw](<https://devfeed.tech/tags/nemoclaw.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-openshell](<https://devfeed.tech/tags/nvidia-openshell.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [openshell](<https://devfeed.tech/tags/openshell.md>), [platform](<https://devfeed.tech/tags/platform.md>), [productivity](<https://devfeed.tech/tags/productivity.md>), [runtime](<https://devfeed.tech/tags/runtime.md>), [secure-ai](<https://devfeed.tech/tags/secure-ai.md>), [secure-ai-agents](<https://devfeed.tech/tags/secure-ai-agents.md>), [security](<https://devfeed.tech/tags/security.md>)

### AI overview

NVIDIA OpenShell provides sandboxed execution and policy-based controls for enterprise AI agents. AIM Intelligence is integrating it with Starfort, whose detection of malicious instructions and data leakage is intended to complement OpenShell's runtime access controls; the company is also considering a preliminary public-sector energy proof of concept. Hyundai Motor Group is exploring NVIDIA NemoClaw blueprints and OpenShell. Separately, LG AI Research is opening API access to K-EXAONE-2.5, developed using NVIDIA Nemotron datasets, including Nemotron-Personas-Korea, a synthetic dataset of 7 million Korean personas grounded in official statistics. The excerpt describes these efforts but does not report proof-of-concept results.

### Source excerpt

As AI agents take on more work, they need capable open models and security that advances alongside them. Hyundai Motor Group is exploring NVIDIA NemoClaw blueprints and the NVIDIA OpenShell runtime, while AIM Intelligence is integrating its Starfort platform with NVIDIA OpenShell to support secure AI agents for employee productivity and enterprise workflows. A key [...]

## Trillion-Parameter RL Weight Syncs Drop From 87 Minutes to 150 Seconds

DevFeed: [Trillion-Parameter RL Weight Syncs Drop From 87 Minutes to 150 Seconds](<https://devfeed.tech/articles/trillion-parameter-rl-weight-syncs-drop-from-87-minutes-to-150-seconds-122231.md>)

Original publisher: [Read original article](<https://hackernoon.com/trillion-parameter-rl-weight-syncs-drop-from-87-minutes-to-150-seconds?source=rss>)

Author: NVIDIA CORPORATION

Published: 2026-11-07T13:00:02Z

Content type: article

Language: en

Sources: [HackerNoon](<https://devfeed.tech/sources/hackernoon.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [NeMo](<https://devfeed.tech/topics/nemo.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [ML model versioning](<https://devfeed.tech/topics/ml-model-versioning.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>)

Tags: [1-5](<https://devfeed.tech/tags/1-5.md>), [1-7](<https://devfeed.tech/tags/1-7.md>), [17](<https://devfeed.tech/tags/17.md>), [2](<https://devfeed.tech/tags/2.md>), [2-2](<https://devfeed.tech/tags/2-2.md>), [2013](<https://devfeed.tech/tags/2013.md>), [2018](<https://devfeed.tech/tags/2018.md>), [2019](<https://devfeed.tech/tags/2019.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [faster](<https://devfeed.tech/tags/faster.md>), [hackernoon-top-story](<https://devfeed.tech/tags/hackernoon-top-story.md>), [llm-infrastructure](<https://devfeed.tech/tags/llm-infrastructure.md>), [making](<https://devfeed.tech/tags/making.md>), [minutes](<https://devfeed.tech/tags/minutes.md>), [model-compression](<https://devfeed.tech/tags/model-compression.md>), [nemo](<https://devfeed.tech/tags/nemo.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [parameter](<https://devfeed.tech/tags/parameter.md>), [reinforcement-learning](<https://devfeed.tech/tags/reinforcement-learning.md>), [rl](<https://devfeed.tech/tags/rl.md>), [rl-weights](<https://devfeed.tech/tags/rl-weights.md>), [syncs](<https://devfeed.tech/tags/syncs.md>), [trillion-parameter-rl](<https://devfeed.tech/tags/trillion-parameter-rl.md>), [trillion-parameter-rl-weight](<https://devfeed.tech/tags/trillion-parameter-rl-weight.md>), [weight](<https://devfeed.tech/tags/weight.md>)

### AI overview

NeMo-DCR accelerates cross-cluster reinforcement-learning weight refits by detecting changed BF16 values and sending only their updates. It maps supported changes into canonical checkpoint coordinates, uses residual conversion for other changes, and encodes updates as XOR masks when the path preserves bits or absolute overwrites otherwise. Receivers apply them through the serving runtime's native loader; overwrite retries repair partial writes, and a joint commit aligns the policy with its source baseline. Payloads stream through object storage or a relay tree without a cross-cluster collective, with delivery pipelined alongside construction and application. In evaluated 30B-1T models at 3% and 5% change rates, refits were 12-40x faster than a transport-only full-checkpoint reference; at 1T and 3%, relay-tree delivery took 150 seconds versus 87.5 minutes. Evaluated BF16 configurations matched dense-refit parameter and buffer bits, and NeMo-DCR runs with receiver failures followed dense NCCL's mean reward and KL trajectories. The latency tests used stress change rates higher than the reported 0.6-1.2% observed across six models.

### Source excerpt

NVIDIA's NeMo-DCR syncs only changed weights, bit-exact, making trillion-parameter RL refits 12-40x faster across clusters.

## The Wizard of Open Weights: Chris Alexiuk Makes AI Easy With Open Models

DevFeed: [The Wizard of Open Weights: Chris Alexiuk Makes AI Easy With Open Models](<https://devfeed.tech/articles/the-wizard-of-open-weights-chris-alexiuk-makes-ai-easy-with-open-models-121053.md>)

Original publisher: [Read original article](<https://blogs.nvidia.com/blog/nvidia-life-chris-alexiuk/>)

Author: Matthew Leib

Published: 2026-11-05T16:00:05Z

Content type: article

Language: en

Sources: [NVIDIA](<https://devfeed.tech/sources/nvidia-blog.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Open Source Models & Datasets](<https://devfeed.tech/topics/open-source-models-datasets.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Open Source](<https://devfeed.tech/topics/open-source.md>), [Learning](<https://devfeed.tech/topics/learning.md>)

Tags: [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [applications](<https://devfeed.tech/tags/applications.md>), [build](<https://devfeed.tech/tags/build.md>), [building](<https://devfeed.tech/tags/building.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [developers](<https://devfeed.tech/tags/developers.md>), [language-models](<https://devfeed.tech/tags/language-models.md>), [large-language-models-llms](<https://devfeed.tech/tags/large-language-models-llms.md>), [llms](<https://devfeed.tech/tags/llms.md>), [models](<https://devfeed.tech/tags/models.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nemotron-labs](<https://devfeed.tech/tags/nemotron-labs.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-life](<https://devfeed.tech/tags/nvidia-life.md>), [open](<https://devfeed.tech/tags/open.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [open-source-ai](<https://devfeed.tech/tags/open-source-ai.md>), [open-weights](<https://devfeed.tech/tags/open-weights.md>)

### AI overview

NVIDIA engineer Chris Alexiuk argues that making AI broadly useful requires openness: access to model weights, training data and recipes. Through YouTube tutorials, he teaches developers to build, fine-tune, evaluate and ship applications using open models. He presents NVIDIA Nemotron as a family with options for different workloads, citing Nemotron 3.5 Lightning for repetitive, high-frequency execution and Nemotron Ultra for long-running autonomous agents, deep research and enterprise tasks. His account emphasizes transparency and accessibility as ways to help more people understand and use AI.

### Source excerpt

Chris Alexiuk has a simple rule about open source AI: If you believe in it, be loud about it. "We should be shouting about this technology from the rooftops," said Alexiuk, a Toronto-based senior product research engineer in deep learning at NVIDIA who specializes in making open large language models (LLMs) practical for working developers. [...]

## Box taps Nvidia's AI models to build a rich context layer for autonomous enterprise agents

DevFeed: [Box taps Nvidia's AI models to build a rich context layer for autonomous enterprise agents](<https://devfeed.tech/articles/box-taps-nvidia-s-ai-models-to-build-a-rich-context-layer-for-autonomous-enterprise-agents-119017.md>)

Original publisher: [Read original article](<https://siliconangle.com/2026/11/05/box-taps-nvidias-ai-models-to-build-a-rich-context-layer-for-autonomous-enterprise-agents/>)

Author: Mike Wheatley

Published: 2026-11-05T14:00:44Z

Content type: news

Language: en

Sources: [SiliconANGLE](<https://devfeed.tech/sources/siliconangle.md>)

Topics: [AI-Ready Data](<https://devfeed.tech/topics/ai-ready-data.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [ai security](<https://devfeed.tech/topics/ai-security.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [aaron-levie](<https://devfeed.tech/tags/aaron-levie.md>), [agent-security](<https://devfeed.tech/tags/agent-security.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-context](<https://devfeed.tech/tags/agentic-context.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent-context](<https://devfeed.tech/tags/ai-agent-context.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-context](<https://devfeed.tech/tags/ai-context.md>), [ai-models](<https://devfeed.tech/tags/ai-models.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [artificial-intelligence-agents](<https://devfeed.tech/tags/artificial-intelligence-agents.md>), [automation](<https://devfeed.tech/tags/automation.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [autonomous-enterprise](<https://devfeed.tech/tags/autonomous-enterprise.md>), [big-data](<https://devfeed.tech/tags/big-data.md>), [box](<https://devfeed.tech/tags/box.md>), [box-architect](<https://devfeed.tech/tags/box-architect.md>), [box-content-retrieval](<https://devfeed.tech/tags/box-content-retrieval.md>), [conference](<https://devfeed.tech/tags/conference.md>), [context](<https://devfeed.tech/tags/context.md>), [context-layer](<https://devfeed.tech/tags/context-layer.md>), [engine](<https://devfeed.tech/tags/engine.md>), [enterprise](<https://devfeed.tech/tags/enterprise.md>), [enterprise-agents](<https://devfeed.tech/tags/enterprise-agents.md>), [mcp-tools](<https://devfeed.tech/tags/mcp-tools.md>), [models](<https://devfeed.tech/tags/models.md>), [nemotron-ultra-3](<https://devfeed.tech/tags/nemotron-ultra-3.md>), [news](<https://devfeed.tech/tags/news.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-models](<https://devfeed.tech/tags/nvidia-models.md>), [open-weight-models](<https://devfeed.tech/tags/open-weight-models.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [siliconangle](<https://devfeed.tech/tags/siliconangle.md>), [the-latest](<https://devfeed.tech/tags/the-latest.md>), [top-story-6](<https://devfeed.tech/tags/top-story-6.md>)

### AI overview

Box announced tools intended to give enterprise AI agents governed access to organizational content and workflows. Box Content Retrieval uses Nvidia Nemotron Retriever embeddings with multimodal and hybrid search to index unstructured files, while retaining native file permissions. Box claims it improves search accuracy by 20%, cuts token use by up to 90%, and speeds responses about fourfold. Box Architect, due in beta later this year, will let workers build workflows in natural language using Box tools and open or closed models. New controls include classification-based file access, MCP guardrails for third-party tools, and real-time alerts for harmful agent activity. Nvidia Nemotron models in Box AI and agent security controls are available now to Enterprise Advanced and Shield customers; Context Retrieval and advanced threat detection are slated for general availability early next year for those subscribers.

### Source excerpt

Few technology firms have embraced the arrival of artificial intelligence agents quite like Box Inc., and today it went even further down the road to automation with an expanded partnership with Nvidia Corp. At its annual customer conference BoxWorks 2026, Box debuted a new retrieval engine that's powered by Nvidia's AI models and designed to [...] The post Box taps Nvidia's AI models to build a rich context layer for autonomous enterprise agents appeared first on SiliconANGLE.

## Box taps Nvidia's Nemotron for new agent-first content retrieval system

DevFeed: [Box taps Nvidia's Nemotron for new agent-first content retrieval system](<https://devfeed.tech/articles/box-taps-nvidia-s-nemotron-for-new-agent-first-content-retrieval-system-119080.md>)

Original publisher: [Read original article](<https://techcrunch.com/2026/11/05/box-taps-nvidias-nemotron-for-new-agent-first-content-retrieval-system/>)

Author: Russell Brandom

Published: 2026-11-05T14:00:00Z

Content type: news

Language: en

Sources: [TechCrunch | AI News & Artificial Intelligence](<https://devfeed.tech/sources/techcrunch-ai-news-artificial-intelligence.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [information retrieval](<https://devfeed.tech/topics/information-retrieval.md>), [search-engines](<https://devfeed.tech/topics/search-engines.md>), [AI Agent](<https://devfeed.tech/topics/ai-agent.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [box](<https://devfeed.tech/tags/box.md>), [box-content-retrieval](<https://devfeed.tech/tags/box-content-retrieval.md>), [embedding](<https://devfeed.tech/tags/embedding.md>), [errors](<https://devfeed.tech/tags/errors.md>), [faster](<https://devfeed.tech/tags/faster.md>), [generative-ai](<https://devfeed.tech/tags/generative-ai.md>), [legal](<https://devfeed.tech/tags/legal.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [tokens](<https://devfeed.tech/tags/tokens.md>)

### AI overview

Box Content Retrieval combines Nvidia's Nemotron Embed models with Box's existing metadata and permission controls so agents can retrieve relevant document chunks instead of reading documents one by one. The concept based embeddings can surface related material even when queries use different wording, with lawyers given as a potential use case for finding clauses and precedents. Nvidia's Kari Briski estimated four times faster answers, 20% fewer errors, and 90% fewer tokens; these are expectations, not reported results. Box said the feature would roll out early next year to Enterprise Advance customers. Although Nemotron's embedding models are open weight and freely available, the system also depends on Box's existing infrastructure, which the excerpt says would make it challenging to recreate from scratch.

### Source excerpt

The system can supposedly answer queries four times faster, with 20% fewer errors and 90% fewer tokens used.

## How NVIDIA Uses AI Agents to Streamline AI Factory Operations

DevFeed: [How NVIDIA Uses AI Agents to Streamline AI Factory Operations](<https://devfeed.tech/articles/how-nvidia-uses-ai-agents-to-streamline-ai-factory-operations-117988.md>)

Original publisher: [Read original article](<https://developer.nvidia.com/blog/how-nvidia-uses-ai-agents-to-streamline-ai-factory-operations/>)

Author: Shara Tibken

Published: 2026-11-04T22:00:17Z

Content type: article

Language: en

Sources: [NVIDIA Developer](<https://devfeed.tech/sources/nvidia-developer.md>), [NVIDIA Technical Blog](<https://devfeed.tech/sources/nvidia-technical-blog.md>)

Topics: [AI in IT operations](<https://devfeed.tech/topics/ai-in-it-operations.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [OpenShell](<https://devfeed.tech/topics/openshell.md>), [AIOps for Kubernetes](<https://devfeed.tech/topics/aiops-for-kubernetes.md>), [control-plane](<https://devfeed.tech/topics/control-plane.md>), [NemoClaw](<https://devfeed.tech/topics/nemoclaw.md>), [Production Engineering](<https://devfeed.tech/topics/production-engineering.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Grafana](<https://devfeed.tech/topics/grafana.md>), [digital sovereignty](<https://devfeed.tech/topics/digital-sovereignty.md>), [DGX Spark](<https://devfeed.tech/topics/dgx-spark.md>)

Tags: [access-control](<https://devfeed.tech/tags/access-control.md>), [agent](<https://devfeed.tech/tags/agent.md>), [agentic-ai-generative-ai](<https://devfeed.tech/tags/agentic-ai-generative-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agent](<https://devfeed.tech/tags/ai-agent.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [ai-factory](<https://devfeed.tech/tags/ai-factory.md>), [build-ai-agents](<https://devfeed.tech/tags/build-ai-agents.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [compute](<https://devfeed.tech/tags/compute.md>), [dashboards](<https://devfeed.tech/tags/dashboards.md>), [data-center-cloud](<https://devfeed.tech/tags/data-center-cloud.md>), [dgx](<https://devfeed.tech/tags/dgx.md>), [engineers](<https://devfeed.tech/tags/engineers.md>), [gb200](<https://devfeed.tech/tags/gb200.md>), [gb300](<https://devfeed.tech/tags/gb300.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [gpu-clusters](<https://devfeed.tech/tags/gpu-clusters.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [hgx](<https://devfeed.tech/tags/hgx.md>), [hpc-scientific-computing](<https://devfeed.tech/tags/hpc-scientific-computing.md>), [machine-learning-artificial-intelligence](<https://devfeed.tech/tags/machine-learning-artificial-intelligence.md>), [mlops](<https://devfeed.tech/tags/mlops.md>), [nemoclaw](<https://devfeed.tech/tags/nemoclaw.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [network](<https://devfeed.tech/tags/network.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [openshell](<https://devfeed.tech/tags/openshell.md>), [operations](<https://devfeed.tech/tags/operations.md>), [reliability](<https://devfeed.tech/tags/reliability.md>), [scale](<https://devfeed.tech/tags/scale.md>), [scheduling](<https://devfeed.tech/tags/scheduling.md>), [site-reliability](<https://devfeed.tech/tags/site-reliability.md>)

### AI overview

NVIDIA's AI Factory Operations Agent correlates signals across compute, network, hardware, and scheduling so SREs can investigate failures through a chat interface. Specialized agents detect issues, analyze cross-layer causes, and recommend actions; examples include job recovery, hardware triage, observability, documentation search, Kubernetes awareness, and cluster management. The framework uses MCP tools and can connect to existing infrastructure integrations. It runs as a Kubernetes control-plane application, with OpenShell creating per-session sandboxes; Kubernetes inspection uses a separate read-only tool with scoped credentials, and the excerpt's example permissions deny access to secrets and pod creation or updates. NVIDIA emphasizes clear agent boundaries, read-only investigation, precise tools, and visible evidence and tool calls. A dedicated GPU-accelerated control plane can isolate orchestration from production workloads, while tenant-specific permissions and sandboxes limit visibility. The framework is experimental open source software, so APIs and deployment requirements may change.

### Source excerpt

As GPU clusters scale, operations span compute, network, hardware, and scheduling. Site reliability engineers (SREs) often have to examine multiple dashboards...

## How NVIDIA Factory Operations Blueprint Coordinates AI Agents to Investigate Production Issues

DevFeed: [How NVIDIA Factory Operations Blueprint Coordinates AI Agents to Investigate Production Issues](<https://devfeed.tech/articles/how-nvidia-factory-operations-blueprint-coordinates-ai-agents-to-investigate-production-issues-117641.md>)

Original publisher: [Read original article](<https://developer.nvidia.com/blog/how-nvidia-factory-operations-blueprint-coordinates-ai-agents-to-investigate-production-issues/>)

Author: Samuel Ochoa

Published: 2026-11-04T17:00:00Z

Content type: article

Language: en

Sources: [NVIDIA Developer](<https://devfeed.tech/sources/nvidia-developer.md>), [NVIDIA Technical Blog](<https://devfeed.tech/sources/nvidia-technical-blog.md>)

Topics: [Multi Agent Systems](<https://devfeed.tech/topics/multi-agent-systems.md>), [agentic workflows](<https://devfeed.tech/topics/agentic-workflows.md>), [Agent Harness](<https://devfeed.tech/topics/agent-harness.md>), [human in the loop agent](<https://devfeed.tech/topics/human-in-the-loop-agent.md>), [Local AI](<https://devfeed.tech/topics/local-ai.md>), [NemoClaw](<https://devfeed.tech/topics/nemoclaw.md>), [OpenShell](<https://devfeed.tech/topics/openshell.md>), [Image processing](<https://devfeed.tech/topics/image-processing.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Isaac](<https://devfeed.tech/topics/isaac.md>)

Tags: [2](<https://devfeed.tech/tags/2.md>), [2026](<https://devfeed.tech/tags/2026.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [actions](<https://devfeed.tech/tags/actions.md>), [adapters](<https://devfeed.tech/tags/adapters.md>), [agent-toolkit](<https://devfeed.tech/tags/agent-toolkit.md>), [agentic-ai-generative-ai](<https://devfeed.tech/tags/agentic-ai-generative-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [blueprint](<https://devfeed.tech/tags/blueprint.md>), [build-ai-agents](<https://devfeed.tech/tags/build-ai-agents.md>), [computer-vision-video-analytics](<https://devfeed.tech/tags/computer-vision-video-analytics.md>), [costs](<https://devfeed.tech/tags/costs.md>), [downtime](<https://devfeed.tech/tags/downtime.md>), [edge-computing](<https://devfeed.tech/tags/edge-computing.md>), [factory](<https://devfeed.tech/tags/factory.md>), [failed](<https://devfeed.tech/tags/failed.md>), [featured](<https://devfeed.tech/tags/featured.md>), [large](<https://devfeed.tech/tags/large.md>), [llms](<https://devfeed.tech/tags/llms.md>), [nemo](<https://devfeed.tech/tags/nemo.md>), [nemoclaw](<https://devfeed.tech/tags/nemoclaw.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [operations](<https://devfeed.tech/tags/operations.md>), [production](<https://devfeed.tech/tags/production.md>), [report](<https://devfeed.tech/tags/report.md>), [robotics](<https://devfeed.tech/tags/robotics.md>), [siemens](<https://devfeed.tech/tags/siemens.md>)

### AI overview

NVIDIA Factory Operations Blueprint (FOX) is a reference architecture for a factory manager agent that connects existing factory systems and delegates work to specialist agents. In the SMT example, an AOI alert prompts a second image inspection; confirmed defects can be sent for rework while a root cause agent uses the board's production history and machine readings to investigate likely causes and produce a sourced report for a line manager. Inconclusive inspections and initial handling require operator review. FOX uses an LLM, an agent harness (NemoClaw by default, with alternatives supported), and NVIDIA OpenShell to isolate execution and enforce access policies. Developers connect MES, ERP, machine databases and procedures; fully on-premises operation requires all models and service endpoints to stay within the factory network, while external models require assessment of data sent to them. Physical robot and AMR handling needs factory-specific integrations; the example demonstrates transport in Isaac Sim and ROS 2. Separately, Siemens describes rush-order simulations through Intelligence Center X and virtual AOI preparation using Process Preparation X, Cosmos and Omniverse, but

### Source excerpt

A Siemens report estimates that unplanned downtime costs an average large plant in the surveyed sectors $253 million per year. Whether investigating a failed...

## When Strong Reasoners Join the Pool: Small Union Premiums and Costly Runaway Generations in LLM Routing

DevFeed: [When Strong Reasoners Join the Pool: Small Union Premiums and Costly Runaway Generations in LLM Routing](<https://devfeed.tech/articles/when-strong-reasoners-join-the-pool-small-union-premiums-and-costly-runaway-generations-in-llm-routing-116897.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.04098>)

Author: Teng-Ruei Chen

Published: 2026-11-04T05:00:00Z

Content type: news

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Model Routing](<https://devfeed.tech/topics/model-routing.md>), [LLM evaluation / benchmarking](<https://devfeed.tech/topics/llm-evaluation-benchmarking.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [math-verify](<https://devfeed.tech/topics/math-verify.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [account](<https://devfeed.tech/tags/account.md>), [all](<https://devfeed.tech/tags/all.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [b300](<https://devfeed.tech/tags/b300.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [budget](<https://devfeed.tech/tags/budget.md>), [but](<https://devfeed.tech/tags/but.md>), [companion](<https://devfeed.tech/tags/companion.md>), [counts](<https://devfeed.tech/tags/counts.md>), [coupling](<https://devfeed.tech/tags/coupling.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decoding](<https://devfeed.tech/tags/decoding.md>), [fixed](<https://devfeed.tech/tags/fixed.md>), [gap](<https://devfeed.tech/tags/gap.md>), [join](<https://devfeed.tech/tags/join.md>), [llm](<https://devfeed.tech/tags/llm.md>), [pool](<https://devfeed.tech/tags/pool.md>), [routing](<https://devfeed.tech/tags/routing.md>), [small](<https://devfeed.tech/tags/small.md>), [union](<https://devfeed.tech/tags/union.md>)

### AI overview

On MATH-500, the measured benefit of taking any correct answer from multiple models was small: the empirical product-coupling premium was 0.215-0.588 percentage points across two- and three-model pools. In a fixed protocol with 30 seeded samples per problem, Nemotron Nano 9B v2 and Nemotron Super 49B v1.5 each reached 96.1% accuracy, versus 76.2% for Llama-3.3-70B-Instruct. Selecting among models offered 1.27 points of empirical headroom over always calling Nano, while sampling Nano three times produced practically the same any-correct rate as calling all three models once. Only 31 responses hit the 65,536-token cap, but those responses contributed up to 60.2% of a model's output-length-squared proxy; long generations also prevented completion of a planned Llama-3.1-8B run. The authors caution that the premium is not directly comparable to a companion study's 3.5-point result because the models, pool size, and protocol differed, and recommend reporting the pool, protocol, samples per cell, and decoding budget with oracle gaps.

### Source excerpt

arXiv:2611.04098v1 Announce Type: new Abstract: Routing benchmarks often judge a router against a hindsight oracle that counts a query as solved whenever any model's recorded answer is correct, but under stochastic decoding that oracle mixes two different objects: the clairvoyant ceiling of committing to one model per query and the premium of calling several models and keeping any correct answer. A companion study measured this premium at 3.5 percentage points (pp) for 11 small open models on MATH-500; how large it is once current reasoning models enter the pool was unknown. We measure it with a fixed-protocol, pre-committed generation on a single NVIDIA B300: all 500 MATH-500 problems, 30 seeded samples per problem, Llama-3.3-70B-Instruct and NVIDIA's Nemotron Nano 9B v2 and Nemotron Super 49B v1.5 (a NAS derivative of the same 70B) under vendor sampling, a 65,536-token cap, a pinned Math-Verify scorer, and an append-only attempt ledger, for 45,000 scored responses. The two reasoning models both reach 96.1% accuracy against 76.2% for the 70B baseline. The empirical product-coupling premium is 0.215-0.588 pp across all two- and three-model pools; the empirical plug-in selection headroom over always calling Nano is 1.27 pp; and resampling Nano three times gives practically the same empirical any-correct rate as calling all three models once. Only 31 responses reached the cap, 25 of them with exactly periodic tails, yet the capped responses account for up to 60.2% of a model's output-length-squared proxy, and long generations made a fourth planned model, Llama-3.1-8B, too slow to finish. The premium is far below the companion's 3.5 pp, but models, pool size and protocol changed together; an oracle gap should therefore be reported with the pool, the protocol, the samples per cell and the decoding budget that produced it.

## Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AI

DevFeed: [Cities Worldwide Build and Scale AI Services With NVIDIA Physical and Agentic AI](<https://devfeed.tech/articles/cities-worldwide-build-and-scale-ai-services-with-nvidia-physical-and-agentic-ai-113599.md>)

Original publisher: [Read original article](<https://blogs.nvidia.com/blog/cities-physical-agentic-ai-applications/>)

Author: Esther Lee

Published: 2026-11-03T07:00:12Z

Content type: article

Language: en

Sources: [NVIDIA](<https://devfeed.tech/sources/nvidia-blog.md>)

Topics: [Agentic AI](<https://devfeed.tech/topics/what-is-agentic-ai.md>), [Omniverse](<https://devfeed.tech/topics/omniverse.md>), [Physical AI](<https://devfeed.tech/topics/physical-ai.md>), [Cosmos](<https://devfeed.tech/topics/cosmos.md>), [Metropolis](<https://devfeed.tech/topics/metropolis.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [digital sovereignty](<https://devfeed.tech/topics/digital-sovereignty.md>), [Critical Infrastructure](<https://devfeed.tech/topics/critical-infrastructure.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-factory](<https://devfeed.tech/tags/ai-factory.md>), [ai-services](<https://devfeed.tech/tags/ai-services.md>), [artificial-intelligence](<https://devfeed.tech/tags/artificial-intelligence.md>), [cameras](<https://devfeed.tech/tags/cameras.md>), [cities](<https://devfeed.tech/tags/cities.md>), [cosmos](<https://devfeed.tech/tags/cosmos.md>), [customer-stories](<https://devfeed.tech/tags/customer-stories.md>), [data](<https://devfeed.tech/tags/data.md>), [digital-twin](<https://devfeed.tech/tags/digital-twin.md>), [digital-twins](<https://devfeed.tech/tags/digital-twins.md>), [events](<https://devfeed.tech/tags/events.md>), [infrastructure](<https://devfeed.tech/tags/infrastructure.md>), [maintenance](<https://devfeed.tech/tags/maintenance.md>), [metropolis](<https://devfeed.tech/tags/metropolis.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-blueprints](<https://devfeed.tech/tags/nvidia-blueprints.md>), [omniverse](<https://devfeed.tech/tags/omniverse.md>), [open-source](<https://devfeed.tech/tags/open-source.md>), [operational](<https://devfeed.tech/tags/operational.md>), [physical-ai](<https://devfeed.tech/tags/physical-ai.md>), [planning](<https://devfeed.tech/tags/planning.md>), [public](<https://devfeed.tech/tags/public.md>), [public-sector](<https://devfeed.tech/tags/public-sector.md>), [scale](<https://devfeed.tech/tags/scale.md>), [sensors](<https://devfeed.tech/tags/sensors.md>), [service](<https://devfeed.tech/tags/service.md>), [services](<https://devfeed.tech/tags/services.md>), [smart-spaces](<https://devfeed.tech/tags/smart-spaces.md>), [sovereign-ai](<https://devfeed.tech/tags/sovereign-ai.md>), [vision](<https://devfeed.tech/tags/vision.md>)

### AI overview

Cities are using shared GPU infrastructure, AI agents, camera analytics and digital twins to extend services across departments while keeping sensitive-data processing under local control. Examples include Spain's 1MillionBot resident-service agents, which handle procedures, appointments, planning incidents and inquiries across more than 150 deployments; Wrocław's shared camera-analysis system for mobility, safety and event response; and Malaysia's managed traffic services, which plan to support 17,000+ cameras by 2027. In tests using Laplace Intelligence's TeraSim, ITMAX reports peak-hour corridor travel times reduced by up to 26%. For planning, Linker Vision and MVRDV report simulating 15-year urban-forest outcomes in 10 minutes, reducing iteration time from two weeks to under an hour. The excerpt also describes sovereign AI infrastructure and startup support, but these initiatives are presented as planned or developing capabilities.

### Source excerpt

Cities generate enormous amounts of data from cameras, sensors and operational systems, but turning it into useful action -- for planning, maintenance and public services -- remains a challenge. Physical and agentic AI offer new ways to close that gap, from digital twins for infrastructure planning to vision and service agents, while keeping local governments [...]

## CISA: Retrofitting Dense Attention Models with Chunk-Indexed Sparse Attention

DevFeed: [CISA: Retrofitting Dense Attention Models with Chunk-Indexed Sparse Attention](<https://devfeed.tech/articles/cisa-retrofitting-dense-attention-models-with-chunk-indexed-sparse-attention-114772.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.03200>)

Author: Mohammed Dabbah, Izhak Golan, Oren Tropp, Daniel Afrimi, Vladimir Anisimov, Tomer Asida, Nave Assaf, Ran El-Yaniv, Ido Galil, Yonatan Geifman, Omri Puny, Ido Shahaf, Akhiad Bercovich

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [Transformer](<https://devfeed.tech/topics/transformer.md>), [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [Cache](<https://devfeed.tech/topics/cache.md>), [NVFP4](<https://devfeed.tech/topics/nvfp4.md>)

Tags: [attention](<https://devfeed.tech/tags/attention.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cisa](<https://devfeed.tech/tags/cisa.md>), [context](<https://devfeed.tech/tags/context.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [decode](<https://devfeed.tech/tags/decode.md>), [gpu-utilization](<https://devfeed.tech/tags/gpu-utilization.md>), [indexing](<https://devfeed.tech/tags/indexing.md>), [inference](<https://devfeed.tech/tags/inference.md>), [llm-inference](<https://devfeed.tech/tags/llm-inference.md>), [models](<https://devfeed.tech/tags/models.md>), [nano](<https://devfeed.tech/tags/nano.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvfp4](<https://devfeed.tech/tags/nvfp4.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-b300](<https://devfeed.tech/tags/nvidia-b300.md>), [prefill](<https://devfeed.tech/tags/prefill.md>), [quality](<https://devfeed.tech/tags/quality.md>), [serving](<https://devfeed.tech/tags/serving.md>), [throughput](<https://devfeed.tech/tags/throughput.md>)

### AI overview

Chunk-Indexed Sparse Attention (CISA) retrofits pretrained dense attention models without gradient updates by selecting fixed-size KV chunks from compact summaries of existing keys. It shares selections across queries within a KV group, while QShare reuses selections across adjacent prefill queries to reduce indexing work. In reported benchmarks, including scoring and selection, CISA was 8.6x faster for late prefill queries and 3.0x faster in decode; core attention alone reached 11.7x and 13.0x. Quality stayed close to dense attention on the named evaluations at the reported chunk budgets, with RULER scores within 2.05 percentage points using 8-16x fewer KV reads. For million-token prompts under the specified four-GPU serving setup, throughput was 2.29x dense attention, rising to 2.61x with fewer chunks; that smaller budget widened the GraphWalks score gap from 1.49 to 3.59 points. Results show a quality-speed tradeoff, and the figures are tied to the stated models, precisions, workloads, and hardware.

### Source excerpt

arXiv:2611.03200v1 Announce Type: new Abstract: Long-context LLM inference is costly because attention repeatedly reads a growing key-value (KV) history. We introduce Chunk-Indexed Sparse Attention (CISA), which retrofits pretrained dense attention models without gradient updates. CISA selects fixed-size KV chunks using compact summaries of existing keys, avoiding token-level indexing and a separate full-length index cache. It uses one indexing query per KV group and shares selected chunks across queries in that group. During prefill, QShare reuses selections across adjacent queries to reduce indexing cost and improve KV reuse and GPU utilization. We evaluate CISA on NVIDIA Nemotron 3 Nano and Ultra, and Inkling-Small. In one-layer BF16 benchmarks at batch size 32 using Ultra's per-device attention dimensions, core attention is 11.7x faster for prefill queries near the end of a 1M-token prompt and 13.0x faster during decode. Including chunk scoring, selection, and all other measured GPU work, speedups are 8.6x and 3.0x. On Ultra NVFP4 with a 4,096-chunk budget, aggregate scores stay within 1.5 percentage points of dense attention on MRCR, LongBench v2, and GraphWalks. RULER sweeps on Nano and Ultra BF16 up to one million tokens show scores within 2.05 percentage points of dense attention with 8-16x fewer KV reads. Serving million-token prompts with 32 concurrent requests and 4K output tokens per request, CISA achieves 2.29x the end-to-end throughput of dense attention using Nemotron 3 Ultra NVFP4 on four NVIDIA B300 GPUs. Selecting fewer chunks increases throughput to 2.61x; in separate GraphWalks evaluations, the smaller budget widens the aggregate score gap from 1.49 to 3.59 percentage points. These results expose a practical quality-speed tradeoff without sparse training.

## How Nvidia's New Nemotron Model Works

DevFeed: [How Nvidia's New Nemotron Model Works](<https://devfeed.tech/articles/how-nvidia-s-new-nemotron-model-works-113899.md>)

Original publisher: [Read original article](<https://blog.bytebytego.com/p/how-nvidias-new-nemotron-model-works>)

Author: ByteByteGo

Published: 2026-11-02T16:31:02Z

Content type: article

Language: en

Sources: [ByteByteGo](<https://devfeed.tech/sources/bytebytego.md>)

Topics: [Nemotron](<https://devfeed.tech/topics/nemotron.md>), [model architecture](<https://devfeed.tech/topics/model-architecture.md>), [LLM optimization](<https://devfeed.tech/topics/llm-optimization.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [NVFP4](<https://devfeed.tech/topics/nvfp4.md>), [grpo](<https://devfeed.tech/topics/grpo.md>), [post-training](<https://devfeed.tech/topics/post-training.md>), [Post-training optimization](<https://devfeed.tech/topics/post-training-optimization.md>)

Tags: [agent](<https://devfeed.tech/tags/agent.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-agents](<https://devfeed.tech/tags/ai-agents.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [attention](<https://devfeed.tech/tags/attention.md>), [autoregressive](<https://devfeed.tech/tags/autoregressive.md>), [cache](<https://devfeed.tech/tags/cache.md>), [cost](<https://devfeed.tech/tags/cost.md>), [costs](<https://devfeed.tech/tags/costs.md>), [model](<https://devfeed.tech/tags/model.md>), [nemotron](<https://devfeed.tech/tags/nemotron.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>)

### AI overview

Nemotron 3 combines Mamba layers, attention, and mixture-of-experts (MoE) layers to reduce per-token computation while retaining some direct access to earlier positions. Mamba carries a fixed-size state instead of growing a key-value cache; attention preserves connections to distant context but still uses KV cache. In the Lightning configuration, 6 of 128 routed experts are selected per token, and 32 query heads share 2 key-value heads. Sparse activation reduces computation, though all experts still contribute to hosting memory. Multi-token prediction can propose several tokens for the main model to verify, reducing sequential generation passes when proposals are accepted. For storage, Nvidia reports using NVFP4 with quantization-aware distillation, reducing a checkpoint from about 66 GB in BF16 to about 22 GB. The article also describes pretraining, supervised fine-tuning, and GRPO reinforcement learning. Its conclusion reports hybrid Mamba-attention and LatentMoE in Nemotron 3 Super and Ultra, with multi-teacher on-policy distillation added for Ultra.

### Source excerpt

In this article, we'll look at how Nemotron works and the various techniques used by Nvidia while building it.

[Next page](<https://devfeed.tech/topics/nemotron.md?cursor=WyIyMDI2LTExLTAyVDE2OjMxOjAyKzAwOjAwIiwgImM2ZTQ4OGViLWExOWYtNGQyZC05MWM2LTZjMjdhNjJhYzEwMyJd>)