# speed Published articles for speed. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Libreboot 20220710 released! DevFeed: [Libreboot 20220710 released!]() Original publisher: [Read original article]() Author: Leah Rowe Published: 2026-09-17T04:32:50.666044Z Content type: release Language: en Sources: [News about Libreboot releases and development]() Topics: [libreboot](), [releases](), [coreboot](), [Linux]() Tags: [article](), [bios](), [build-system](), [canoeboot](), [coreboot](), [documentation](), [free-software](), [gcc](), [hardware](), [libre](), [libreboot](), [linux](), [opensource](), [release](), [releases](), [scripts](), [security](), [speed](), [testing](), [uefi](), [upgrade](), [usb](), [version]() ### AI overview Libreboot 20220710 is a stable bug-fix release relative to version 20211122. It adds documentation snapshots and fixes or improves GRUB boot performance, laptop compatibility, build behavior, USB keyboard support, coreboot configuration, and related tooling. No new boards or major features were added. ### Source excerpt Article: Libreboot 20220710 released! Web link: https://libreboot.org/news/libreboot20220710.html ## How fast is C++23's std::flat\_map? DevFeed: [How fast is C++23's std::flat\_map?]() Original publisher: [Read original article]() Author: Daniel Lemire Published: 2026-09-16T20:26:36Z Content type: article Language: en Sources: [Daniel Lemire]() Topics: [Data structures](), [Library](), [gcc](), [LLVM](), [clang]() Tags: [array](), [arrays](), [clang](), [gcc](), [llvm](), [overhead](), [speed](), [standard-library]() ### AI overview This article benchmarks C++23's std::flat_map, a sorted container backed by parallel arrays of keys and values. It explains serialization and loading considerations, then compares insertion and lookup performance with std::map. Random-order insertion becomes quadratic as the container grows, while increasing-order or bulk insertion is much faster; random lookups can also be faster for large maps because std::flat_map uses less memory. ### Source excerpt C++23 added a new type to the standard library: std::flat_map. There is also a std::flat_set and other variants, but let me focus on std::flat_map. A flat map is a sorted vector of keys next to a vector of values. A query is a binary search over the sorted keys. You need a recent standard library: ... Continue reading How fast is C++23's std::flat_map? ## Micron Shows off 512GB DDR5 RDIMM: 12TB per Dual-Socket Server at 9,200 MT/s, Volume Production in 2H 2027 DevFeed: [Micron Shows off 512GB DDR5 RDIMM: 12TB per Dual-Socket Server at 9,200 MT/s, Volume Production in 2H 2027]() Original publisher: [Read original article]() Author: Brian Beeler Published: 2026-09-15T20:18:17Z Content type: news Language: en Sources: [StorageReview.com]() Topics: [ddr5](), [servers](), [intel](), [cpu]() Tags: [capacity](), [ddr5](), [enterprise](), [generation](), [intel](), [memory](), [modules](), [performance](), [production](), [release](), [server](), [speed](), [volume]() ### AI overview Micron demonstrated a 512GB DDR5 RDIMM rated for up to 9,200 MT/s. The module can provide 12TB of memory in a 24-slot dual-socket server, with volume production scheduled for the second half of 2027. AMD and Intel are validating it for next-generation server platforms. ### Source excerpt Micron has demonstrated a 512GB DDR5 RDIMM running on multiple server platforms, which it calls the world's first module at that capacity, and says AMD and Intel are both validating it for their next-generation server platforms. The module is rated for speeds up to 9,200 MT/s, and in a 24-slot dual-socket server it puts 12TB The post Micron Shows off 512GB DDR5 RDIMM: 12TB per Dual-Socket Server at 9,200 MT/s, Volume Production in 2H 2027 appeared first on StorageReview.com. ## How energy teams turn theft detection into governed action with Genie and AI business processes DevFeed: [How energy teams turn theft detection into governed action with Genie and AI business processes]() Original publisher: [Read original article]() Author: Daniel Zoccali; Jack Yallop Published: 2026-09-15T16:50:00Z Content type: article Language: en Sources: [Databricks]() Topics: [databricks](), [Machine learning](), [Artificial Intelligence]() Tags: [databricks](), [energy](), [fraud](), [industries](), [ml](), [model](), [reporting](), [safety](), [speed]() ### AI overview The article explains how energy teams can operationalize energy-theft detection by connecting model-generated risk signals with investigation, field operations, revenue recovery, and reporting in a governed workflow. It presents a Databricks implementation using a Databricks App, Lakebase, and Unity Catalog. ### Source excerpt Energy theft is the deliberate use of gas or electricity without paying for it, typically... ## Zstd Improvement For Linux 7.4 To Avoid Redundant Initialization DevFeed: [Zstd Improvement For Linux 7.4 To Avoid Redundant Initialization]() Original publisher: [Read original article]() Author: Michael Larabel Published: 2026-09-13T14:04:23Z Content type: news Language: en Sources: [Phoronix]() Topics: [Compression](), [Linux](), [Benchmark](), [Cryptography]() Tags: [benchmark](), [compression](), [crypto](), [desktop-linux](), [development](), [kernel](), [linux](), [linux-benchmarking](), [linux-hardware-benchmarks](), [linux-hardware-reviews](), [linux-how-to](), [linux-performance](), [linux-server-benchmarks](), [open-source-graphics](), [performance](), [phoronix](), [phoronix-test-suite](), [speed](), [ubuntu-benchmarks](), [ubuntu-hardware]() ### AI overview The article reports Linux 7.4 patches that defer Zstd stream initialization until the first walk iteration, eliminating redundant initialization. Benchmarks show single-digit compression speed improvements and decompression speedups of 13% on bare metal or 35% in a virtual machine. ### Source excerpt In addition to Usama Arif's recent Linux patches for addressing a major inefficiency within the Linux kernel's Zstd compression code, he also has a separate patch series destined for Linux 7.4 to further enhance the Zstd compression/decompression performance by avoiding redundant initialization... ## Mold High Speed Linker Being Rewritten In Rust, Hopes To Be The Default Linker On Linux DevFeed: [Mold High Speed Linker Being Rewritten In Rust, Hopes To Be The Default Linker On Linux]() Original publisher: [Read original article]() Author: Michael Larabel Published: 2026-09-11T13:21:42Z Content type: news Language: en Sources: [Phoronix]() Topics: [Rust](), [Linux](), [Programming](), [toolchain]() Tags: [compatibility](), [desktop-linux](), [firmware](), [kernels](), [linux](), [linux-benchmarking](), [linux-hardware-benchmarks](), [linux-hardware-reviews](), [linux-how-to](), [linux-performance](), [linux-server-benchmarks](), [open-source](), [open-source-graphics](), [performance](), [phoronix](), [phoronix-test-suite](), [rust](), [speed](), [ubuntu-benchmarks](), [ubuntu-hardware]() ### AI overview The Mold high-speed linker is being rewritten from C++ in Rust for Mold 3.0. Its goals include preserving Mold 2.x performance, improving safety, supporting missing linker-script features, and potentially adding incremental linking. ### Source excerpt The Mold high performance linker that typically outperforms the likes of LLVM lld, GNU LD, Wild, and others is being rewritten in the Rust programming language and also has an ambitious goal of being adopted as the default linker on Linux systems... ## tsgolint Reaches Stable v7, Bringing Go-Powered Type-Aware Linting to Oxlint DevFeed: [tsgolint Reaches Stable v7, Bringing Go-Powered Type-Aware Linting to Oxlint]() Original publisher: [Read original article]() Author: Daniel Curtis Published: 2026-09-11T12:02:00Z Content type: news Language: en Sources: [InfoQ]() Topics: [ESLint](), [Compiler]() Tags: [benchmarks](), [development](), [eslint](), [go-language](), [javascript](), [news](), [performance](), [release](), [speed](), [tsgolint-oxlint-typescript](), [typescript](), [web-development]() ### AI overview tsgolint v7 is now stable, adding Go-powered, type-aware TypeScript linting to Oxlint. It supports 59 of 61 typescript-eslint type-aware rules and is reported as 12 to 18 times faster than ESLint in the cited benchmarks. ### Source excerpt tsgolint has released a stable v7, enhancing TypeScript linting with native Go speed. It offers type-aware linting, leveraging TypeScript's semantic analysis through the typescript-go compiler. Oxlint manages configurations and file discovery. The release, compatible with TypeScript 7.0.2, handles 59 of 61 type-aware rules and shows significant performance improvements over ESLint. By Daniel Curtis ## Oracle says AI will save it from the SaaSpocalypse, not bring it on DevFeed: [Oracle says AI will save it from the SaaSpocalypse, not bring it on]() Original publisher: [Read original article]() Author: Simon Sharwood Published: 2026-09-11T02:43:10Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [AI, ML & Data Engineering]() Tags: [ai](), [ai-and-ml](), [applications](), [oracle](), [saas](), [sales](), [software](), [speed]() ### AI overview Oracle says AI can provide a better interface, speed installations, and drive IaaS sales rather than cause a SaaS decline. ### Source excerpt It's a better interface, can speed installations, and drives IaaS sales too ## Redefining Digital Resilience in the AI Era DevFeed: [Redefining Digital Resilience in the AI Era]() Original publisher: [Read original article]() Author: Kamal Hathi Published: 2026-09-10T17:00:15Z Content type: news Language: en Sources: [Cisco Blogs]() Topics: [Resilience](), [Artificial Intelligence](), [data](), [Architecture & Design](), [Security](), [SOC](), [monitor]() Tags: [agentic-ai-security](), [agentic-ops](), [agentic-security](), [ai](), [ai-agents](), [autonomous](), [cost](), [digital-resilience](), [executive-platform](), [governance](), [monitor](), [observability](), [resilience](), [scale](), [security](), [soc](), [speed](), [workflows]() ### AI overview Splunk is evolving its data platform for the AI era, focusing on digital resilience, trusted autonomous agents, machine-speed security operations, and economically governed AI workflows. ### Source excerpt Splunk is evolving for the AI era - removing constraints to help customers achieve digital resilience and leverage AI with speed, scale and trust. ## Adaptive Instructed-Retriever: Frontier-Quality Search at 2x Lower Latency DevFeed: [Adaptive Instructed-Retriever: Frontier-Quality Search at 2x Lower Latency]() Original publisher: [Read original article]() Author: Cindy Wang; Cheng Li; Jialu Liu; Sean Kulinski; Arnav Singhvi; Wen Sun; Michael Bendersky Published: 2026-09-09T13:30:00Z Content type: article Language: en Sources: [Databricks]() Topics: [databricks](), [Latency](), [Low Latency](), [data]() Tags: [agents](), [announcements](), [benchmarks](), [cost](), [data](), [databricks](), [enterprise](), [latency](), [low-latency](), [performance](), [platform](), [retrieval](), [search](), [speed](), [third-party]() ### AI overview Databricks introduces Adaptive Instructed-Retriever, a retrieval model that combines fast parallel search with sequential multi-step search for harder enterprise queries. It adaptively spends extra computation only when useful, achieving comparable quality to leading third-party models at twice lower latency while improving over single-step retrieval on reported benchmarks. ### Source excerpt Effective enterprise data agents require search that is both accurate and fast. Earlier... ## Introducing ChatGPT Images 2.5 DevFeed: [Introducing ChatGPT Images 2.5]() Original publisher: [Read original article]() Published: 2026-09-08T11:30:00Z Content type: release Language: en Sources: [OpenAI News]() Topics: [API]() Tags: [api](), [chatgpt](), [codex](), [feature](), [generate](), [generation](), [gpt](), [images](), [latency](), [mobile](), [models](), [product](), [speed](), [web](), [work](), [workflows]() ### AI overview ChatGPT Images 2.5 is a new image model with sharper details, more precise multi-turn editing, improved reference-photo fidelity, richer textures, and up to 50% lower generation latency than Images 2.0. The release adds creative controls in ChatGPT and introduces GPT-Image-2.5 Flare and Sunburst in the API. ### Source excerpt ChatGPT Images 2.5 helps turn your ideas, sketches, and reference photos into more personalized, polished images that better reflect your ideas. ## I Tested the Best Wi-Fi Adapters for Raspberry Pi DevFeed: [I Tested the Best Wi-Fi Adapters for Raspberry Pi]() Original publisher: [Read original article]() Author: Patrick Fromaget Published: 2026-09-05T05:00:00Z Content type: comparison Language: en Sources: [RaspberryTips]() Topics: [Hardware](), [USB](), [Linux]() Tags: [comparison](), [cost](), [hardware](), [inspiration](), [linux](), [raspberry-pi](), [raspberry-pi-5](), [router](), [speed]() ### AI overview A hands-on comparison of USB Wi-Fi adapters for Raspberry Pi, evaluating speed, signal strength, stability, Linux support, cost, and bulk against the Raspberry Pi 5's built-in Wi-Fi. The article highlights the BrosTrend AXE3000 as the best overall option, the TP-Link Archer T3U Plus as the best value, and notes that Raspberry Pi 5 users may not need an external adapter. ### Source excerpt Over the years, I've mostly used the default Wi-Fi on my Raspberry Pis, and it's good enough for most projects. But when you start putting the Pi in a case, moving it from the router, or using a fancy Wi-Fi access point that can handle faster speeds, it makes sense to look into better solutions.... ## The Agentic Analytics Benchmark: Measuring model accuracy and efficiency in analytical agents DevFeed: [The Agentic Analytics Benchmark: Measuring model accuracy and efficiency in analytical agents]() Original publisher: [Read original article]() Author: Eduardo Vellasques; Al Brown Published: 2026-09-01T19:15:00Z Content type: article Language: en Sources: [ClickHouse Blog]() Topics: [LLM evaluation / benchmarking](), [benchmarking](), [text2sql](), [Databases](), [Claude](), [deepseek]() Tags: [agentic](), [agents](), [benchmark](), [claude](), [cost](), [deepseek](), [model](), [speed](), [sql]() ### AI overview The article introduces data-agent-mnist, an open benchmark harness for evaluating analytics agents against a data warehouse. It compares 29 models on correctness, cost, speed, and number of turns, and contrasts agentic analytics with conventional text-to-SQL evaluation. ### Source excerpt We took 201 real analytics questions from our production data warehouse, benchmarked 29 models on correctness, cost, and speed, and released an open harness so you can run the same test on your own. ## IBM Quantum Nighthawk r2--more circuits, faster DevFeed: [IBM Quantum Nighthawk r2--more circuits, faster]() Original publisher: [Read original article]() Published: 2026-08-31T14:30:00Z Content type: article Language: en Sources: [IBM Research]() Topics: [ibm](), [Quantum Computing](), [Hardware]() Tags: [hardware](), [ibm](), [performance](), [processor](), [quantum](), [quantum-computing](), [quantum-error-correction-mitigation](), [quantum-systems](), [release](), [speed]() ### AI overview IBM Quantum Nighthawk r2 uses independent high-speed qubit reset to execute more than 100,000 circuits per second, providing up to 25 times the circuit throughput of IBM Quantum Heron. The 120-qubit processor also supports accurate results on circuits with more than 7,500 gates. ### Source excerpt High-speed, independent qubit reset boosts circuit throughput 25x over Heron while enabling accurate observable estimation on circuits with 7,500+ gates. ## The Custodial Era of UX: Cleaning Up After AI DevFeed: [The Custodial Era of UX: Cleaning Up After AI]() Original publisher: [Read original article]() Author: Anna Kaley, Raluca Budiu Published: 2026-08-28T17:00:00Z Content type: article Language: en Sources: [NN/g latest articles and announcements]() Topics: [User experience (UX)](), [Artificial Intelligence](), [Usability](), [Accessibility](), [Information Architecture](), [User interface design]() Tags: [accessibility](), [ai](), [article](), [cognitive-load](), [complexity](), [information-architecture](), [interaction-design](), [mental-models](), [speed](), [usability](), [ux]() ### AI overview AI enables teams to create content, prototypes, code, and working features faster than UX teams can evaluate them, creating UX debt. The article describes UX's role in reviewing AI-generated work, assessing its usefulness and usability, and restoring clarity, accessibility, trust, and coherence. ### Source excerpt AI lets teams build faster than UX can evaluate. UX can adapt by building shared judgment, accelerating evaluation, and guiding AI-generated designs. ## This custom motorized turntable makes welding easier DevFeed: [This custom motorized turntable makes welding easier]() Original publisher: [Read original article]() Author: Arduino Team Published: 2026-08-28T12:14:44Z Content type: article Language: en Sources: [Arduino Blog]() Topics: [Arduino]() Tags: [arduino](), [electronics](), [motor](), [motorized-turntable](), [rotary-table](), [rotation](), [speed](), [uno](), [welding](), [welding-table]() ### AI overview Greenhill Forge built an Arduino-controlled rotary welding table to rotate round workpieces at an adjustable, consistent speed. The design uses an Arduino UNO Rev3, a stepper motor with a gear reducer, a rotary encoder, and an OLED screen, while a ground strap provides a current path that keeps welding current away from the electronics. ### Source excerpt MIG welding isn't very difficult -- at least not until you need to weld something at a weird angle or over a long seam. Those situations will make your welds look like they were done by an inebriated amateur. The best way to handle that is by setting yourself up to maintain stability through the [...] The post This custom motorized turntable makes welding easier appeared first on Arduino Blog. ## Speed Insights now has a free tier DevFeed: [Speed Insights now has a free tier]() Original publisher: [Read original article]() Author: Chris Widmaier Published: 2026-08-25T16:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [Vercel](), [dashboards](), [Command-line interface](), [data]() Tags: [cli](), [free](), [performance](), [product](), [speed](), [time](), [upgrade](), [vercel]() ### AI overview Vercel Speed Insights now offers a free tier on every plan for any number of projects, including 10,000 events per team every 30 days. The paid tier is now called Speed Insights Plus and adds deeper diagnostics, historical data, Drains, and Vercel CLI access. Existing paid customers keep the same capabilities and price. ### Source excerpt Speed Insights now has a free tier that gives you a high-level performance overview from your real users. The new free tier: Is available on every plan, for any number of projects Includes 10,000 events per team, every 30 days Install Speed Insights: Previously, free Speed Insights was limited to a single project on Hobby, and upgrading to Pro meant losing access unless you paid for the add-on. Now your free tier carries over, and you only pay if you upgrade. The paid product, now called Speed Insights Plus, includes deeper diagnostics and historical data. If you were already paying for Speed Insights, only the name has changed, and your capabilities and price are the same. Install @vercel/speed-insights in your project to start reporting real-user data, or open the Speed Insights dashboard to view your results. Learn more in the Speed Insights documentation. Speed Insights reports how real visitors experience your pages, using live traffic rather than lab conditions. Speed Insights Plus adds slow-route diagnosis and detailed breakdowns for finding where pages get slow, and historical data to track performance over time. It also provides access to Drains and the Vercel CLI for working with performance data outside the dashboard. Read more ## Jalapeño's first results show industry-leading speed and efficiency in AI inference DevFeed: [Jalapeño's first results show industry-leading speed and efficiency in AI inference]() Original publisher: [Read original article]() Published: 2026-08-25T07:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [Artificial Intelligence](), [Inference](), [Inference Performance](), [OpenAI](), [Hardware](), [Benchmark](), [gpt-oss](), [deepseek]() Tags: [ai](), [benchmark](), [deepseek](), [efficiency](), [engineering](), [gpt-oss](), [hardware](), [inference](), [latency](), [models](), [openai](), [performance](), [software](), [speed](), [systems]() ### AI overview OpenAI reports initial results for Jalapeño, its custom inference chip. The article says the chip delivers higher throughput, lower end-to-end latency, and greater AI work per watt across GPT-OSS 120B, DeepSeek R1, and Kimi K2.5 1T, based on tests using the InferenceX benchmark. ### Source excerpt Jalapeño is a custom inference chip from OpenAI that delivers faster, more power-efficient AI inference, with higher throughput and lower latency for modern models. ## How NVIDIA Groq 3 LPX Unlocks Ultrafast Interactivity at Long Context on NVIDIA Vera Rubin DevFeed: [How NVIDIA Groq 3 LPX Unlocks Ultrafast Interactivity at Long Context on NVIDIA Vera Rubin]() Original publisher: [Read original article]() Author: Tanya Lenz Published: 2026-08-24T15:00:00Z Content type: article Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [d-matrix](), [Vera Rubin](), [Inference Performance](), [long-context](), [Benchmark](), [Artificial Intelligence](), [gemma4](), [systems](), [Cache]() Tags: [agent](), [agentic](), [agentic-ai-generative-ai](), [ai](), [ai-factory](), [ai-inference](), [benchmark](), [cache](), [collective](), [data-center-cloud](), [developer-tools-techniques](), [groq](), [groq-3-lpx](), [inference-performance](), [long-context](), [low-latency-inference](), [lpx](), [nvidia](), [nvidia-vera](), [performance](), [rubin-gpu](), [speed](), [systems](), [tokens](), [training-ai-models](), [vera-rubin](), [vera-rubin-nvl72]() ### AI overview NVIDIA Groq 3 LPX, paired with Vera Rubin NVL72, delivers high-interactivity AI inference for long-context workloads. A reported benchmark measured 3,431 output tokens per second on Gemma 4 31B with a 100K context. ### Source excerpt NVIDIA Groq 3 LPX is the interactive AI inference accelerator for the NVIDIA Vera Rubin platform. At the core of the platform is NVIDIA Vera Rubin NVL72, the... ## What is a Minimum Viable Platform (MVP)? DevFeed: [What is a Minimum Viable Platform (MVP)?]() Original publisher: [Read original article]() Author: Luca Galante Published: 2026-08-21T14:37:28Z Content type: article Language: en Sources: [Platform Engineering Blog]() Topics: [Platform Engineering](), [internal developer platform](), [Developer Platform](), [Shared Responsibility Model]() Tags: [engineering](), [evaluation](), [gartner](), [idp](), [internal-developer-platform](), [organizational](), [platform](), [platform-engineering](), [shared-responsibility](), [speed]() ### AI overview This article explains why platform engineering initiatives often fail despite widespread adoption of internal developer platforms. It argues that culture, process, and delivery speed matter more than tooling choices, and recommends starting with a Minimum Viable Platform to build stakeholder support and deliver value quickly. ### Source excerpt Unlock success in platform engineering: Avoid common pitfalls, embrace MVP approach, and win stakeholder support for a transformative platform initiative. ## Open-weight models are fast on Neon AI Gateway. Here's why DevFeed: [Open-weight models are fast on Neon AI Gateway. Here's why]() Original publisher: [Read original article]() Author: Carlota Soto Published: 2026-08-20T12:00:00Z Content type: article Language: en Sources: [Blog -- Neon Docs]() Topics: [databricks](), [Inference Performance](), [model-deployment](), [Low-Latency Inference](), [Caching](), [Mixture of Experts (MoE)](), [gpt-oss](), [Large Language Model]() Tags: [caching](), [databricks](), [gpt-oss](), [inference](), [mixture-of-experts-moe](), [optimization](), [performance](), [product](), [speed]() ### AI overview The article explains why open-weight models can deliver low latency on Neon AI Gateway when paired with Databricks Foundation Model APIs and an optimized serving stack. It highlights continuous batching, KV-cache paging, custom kernels, quantization, multi-GPU layouts, prompt caching, and Mixture-of-Experts architectures. ### Source excerpt Open-weight models are built for performance, but how they're served determines whether end users actually feel that speed. Neon AI Gateway calls Databricks Foundation Model APIs for Databricks-hosted open-weight models, a stack that has years of inference engineering behind it. ## Why global workers are driving demand for stablecoin payouts DevFeed: [Why global workers are driving demand for stablecoin payouts]() Original publisher: [Read original article]() Author: Siddharth Kulkarni; Mitchell Martins-Collum Published: 2026-08-19T00:00:00Z Content type: article Language: en Sources: [Stripe Blog]() Topics: [stripe](), [Transactions](), [Meta]() Tags: [cost](), [cross-border](), [fees](), [global](), [payments](), [speed](), [stripe](), [transactions]() ### AI overview This article examines why global gig workers, contractors, creators, freelancers, and marketplace sellers are increasingly interested in stablecoin payouts. Drawing on a Stripe survey of more than 2,300 workers across 20 countries, it highlights dissatisfaction with the fees and slow settlement of traditional cross-border payments, as well as demand for inflation protection, financial access, remittance relief, rewards on balances, education, and simpler wallet setup. ### Source excerpt Platforms like DoorDash, Meta, and Deel already enable stablecoin payouts for global workers. We surveyed 2,300 workers in 20 countries to see what's driving stablecoin demand, where the opportunity is highest, and how other platforms can adapt. ## Developing Nemotron 3.5 Lightning NVFP4 with QAD Using NVIDIA Model Optimizer DevFeed: [Developing Nemotron 3.5 Lightning NVFP4 with QAD Using NVIDIA Model Optimizer]() Original publisher: [Read original article]() Author: Tanya Lenz Published: 2026-08-17T18:12:48Z Content type: tutorial Language: en Sources: [NVIDIA Developer](), [NVIDIA Technical Blog]() Topics: [Nemotron](), [NVFP4](), [quantization](), [Post-training optimization](), [Nvidia](), [post-training](), [Mamba]() Tags: [agentic-ai-generative-ai](), [benchmarks](), [compute](), [developer-tools-techniques](), [developers](), [edge-computing](), [evaluation](), [featured](), [latency](), [mamba](), [megatron](), [memory](), [model](), [model-optimizer](), [models](), [nemotron](), [nvfp4](), [nvidia](), [open](), [post-training](), [quantization](), [speed](), [training](), [training-ai-models]() ### AI overview This tutorial explains how quantization-aware distillation (QAD) creates the Nemotron 3.5 Lightning NVFP4 checkpoint using NVIDIA Model Optimizer. It covers post-training quantization, teacher-student distillation, and evaluation, showing how QAD can recover accuracy while reducing memory usage and increasing throughput. ### Source excerpt Teams customize their models to hit their targets for latency, speed, memory, and compute. With the open NVIDIA Nemotron family of models, developers can find... ## Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things DevFeed: [Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-08-16T22:00:39Z Content type: opinion Language: en Sources: [Simon Willison]() Topics: [qwen](), [Large Language Model](), [Hardware](), [DGX Spark](), [NVIDIA DGX](), [SVG]() Tags: [ai](), [ai-2-235](), [ai-in-china](), [ai-in-china-108](), [coding-agents](), [coding-agents-248](), [cost](), [dgx-spark](), [generative-ai](), [generative-ai-1-981](), [llama-cpp](), [llama-cpp-29](), [llm](), [llm-reasoning](), [llm-reasoning-103](), [llm-release](), [llm-release-231](), [llms](), [llms-1-947](), [lm-studio](), [lm-studio-23](), [local-llms](), [local-llms-164](), [nvidia-dgx](), [nvidia-spark](), [nvidia-spark-6](), [pelican-riding-a-bicycle](), [pelican-riding-a-bicycle-142](), [pi](), [pi-6](), [qwen](), [qwen-61](), [reasoning](), [speed](), [svg]() ### AI overview Simon Willison evaluates Qwen 3.8 27B, a vision-capable 27-billion-parameter LLM that can run locally on suitable hardware. He finds that its default xhigh reasoning setting consumes substantial context and time, while adjusting the reasoning effort and increasing the context limit improves practicality. He also reports strong results generating an SVG locally. ### Source excerpt Friday's big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba's Qwen research lab. I've been looking forward to this one: 27B is an excellent size for running a model on a reasonably specced laptop, and its predecessor Qwen 3.6 27B was impressive. Qwen's self-reported benchmarks for this model are eye-opening. They show a boost from both Qwen 3.6 27B and the closed-weight Qwen 3.7-Plus, which was one of Qwen's strongest models of any size as recently as May this year. It will be interesting to hear what independent benchmarks have to say about the model. I've been running the model on two different machines: my 128GB M5 Max MacBook Pro, and an NVIDIA DGX Spark. On both machines I'm running LM Studio and their 17GB Q4_K_M quantized build. I also tried using llama-server directly on the Spark. The default of extra high results in spectacular over-thinking Qwen's documentation describes the model as defaulting to xhigh for the reasoning effort, and the LM Studio GGUF I've been trying preserves that default: Qwen3.8 comes with official support for reasoning_effort, which can be used to adjust reasoning depth and control cost: xhigh (default): for complex tasks demanding thorough analysis medium: balancing accuracy and speed low: efficient reasoning optimizing for speed and cost This is a hilarious default. It's absolutely not a good way to run the model, especially on consumer hardware. I've been finding the results extremely entertaining. I quickly ran into problems with LM Studio's default context limit of 8,192 tokens - Qwen was using them all up thinking about even the most mundane of problems. I loaded the model with the full 262,144 maximum context length and that problem went away. Here's the pelican riding a bicycle SVG I got from my first attempt with that increased context length. It took 21 minutes to generate, using 22,276 reasoning tokens to produce 3,223 tokens of output. You can read the reasoning trace here. Th [Next page]()