# model Published articles for model. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## AI model watermarking changes agent behavior DevFeed: [AI model watermarking changes agent behavior]() Original publisher: [Read original article]() Author: Thomas Claburn Published: 2026-09-17T13:00:00Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [watermarking](), [Artificial Intelligence]() Tags: [agent](), [ai](), [ai-and-ml](), [ai-watermarking](), [anthropic](), [model](), [prompt-injection](), [synthid](), [watermarking]() ### AI overview Lasso Security reports that AI model watermarking is associated with differences in agent tool handling and model refusals. ### Source excerpt Lasso Security sees differences in tool handling and model refusals ## Our framework for reporting model misalignment DevFeed: [Our framework for reporting model misalignment]() Original publisher: [Read original article]() Published: 2026-09-16T17:00:00Z Content type: article Language: en Sources: [OpenAI News]() Topics: [OpenAI](), [AI Development](), [AI Strategy](), [Framework](), [Development]() Tags: [ai](), [ai-development](), [alignment](), [behavior](), [developers](), [model](), [monitoring](), [openai](), [reporting](), [research](), [standards]() ### AI overview OpenAI introduces a framework for tracking, investigating, and disclosing model misalignment, accompanied by six reports on unexpected or concerning model behavior observed over the previous six months. The framework favors disclosure even when the significance of an instance is uncertain and is intended to evolve through experience and public feedback. ### Source excerpt OpenAI shares a framework for tracking, investigating, and disclosing model misalignment, alongside six reports of unexpected or concerning model behavior. ## New AI technique could make minimally invasive surgeries safer and more precise DevFeed: [New AI technique could make minimally invasive surgeries safer and more precise]() Original publisher: [Read original article]() Author: Adam Zewe | MIT News Published: 2026-09-16T15:00:00Z Content type: news Language: en Sources: [MIT AI News]() Topics: [Artificial Intelligence](), [AI Research](), [3D](), [navigation]() Tags: [3d](), [ai](), [artificial-intelligence](), [computer-science-and-artificial-intelligence-laboratory-csail](), [computer-science-and-technology](), [computer-vision](), [electrical-engineering-and-computer-science-eecs](), [health-care](), [images](), [imaging](), [jameel-clinic](), [machine-learning](), [medical-devices](), [medical-imaging](), [minimally-invasive-surgery](), [mit-ibm-computing-research-lab](), [mit-schwarzman-college-of-computing](), [model](), [national-institutes-of-health-nih](), [navigation](), [paper](), [polina-golland](), [precision](), [real-time](), [research](), [school-of-engineering](), [vision](), [vivek-gopalakrishnan]() ### AI overview MIT researchers and collaborators developed xvr, an AI method that adapts to individual patients and rapidly aligns intraoperative X-rays with preoperative 3D medical scans. The technique is intended to improve surgical navigation for minimally invasive procedures. ### Source excerpt This patient-specific method, called xvr, helps doctors use X-rays for surgical navigation in fields such as orthopedics and neurosurgery. ## Arduino announces a live build of a privacy-focused smart doorbell on the Arduino UNO Q DevFeed: [Arduino announces a live build of a privacy-focused smart doorbell on the Arduino UNO Q]() Original publisher: [Read original article]() Author: Arduino Team Published: 2026-09-16T14:06:15Z Content type: article Language: en Sources: [Arduino Blog]() Topics: [Arduino](), [Computer vision](), [Security]() Tags: [arduino](), [computer-vision](), [model](), [notify](), [privacy](), [security](), [smart-doorbell](), [uno-q]() ### AI overview Arduino announces a live build showing how to create a smart doorbell using a computer vision model that runs locally on an Arduino UNO Q board. The event is scheduled for September 22 at 3 PM CET / 9 AM ET and will include questions for the Arduino team. ### Source excerpt Go on your favorite online shopping platform, and you'll find any number of smart doorbell options. Click to purchase, have it delivered, install it, download some app. But where's the fun in that? And also, don't you wonder how that thing works? That thing that watches you and your loved ones go in and out, [...] The post Build your own smart doorbell and protect your privacy - in one hour, with Massimo Banzi appeared first on Arduino Blog. ## TypeSafe AI debuts model for machines that plays Doom DevFeed: [TypeSafe AI debuts model for machines that plays Doom]() Original publisher: [Read original article]() Author: Thomas Claburn Published: 2026-09-16T01:35:16Z Content type: release Language: en Sources: [www.theregister.com - Articles]() Topics: [Artificial Intelligence](), [ai and ml]() Tags: [ai](), [ai-agents](), [ai-and-ml](), [artificial-intelligence](), [jev](), [large-language-models](), [model](), [typesafe-ai]() ### AI overview TypeSafe AI introduced Jev, a model that produces typed probabilistic decisions rather than conversational responses. ### Source excerpt 'Jev' doesn't chat. It produces typed probabilistic decisions ## TypeSafe AI's Jev now available on AI Gateway DevFeed: [TypeSafe AI's Jev now available on AI Gateway]() Original publisher: [Read original article]() Author: Jerilyn Zheng Published: 2026-09-16T00:00:00Z Content type: release Language: en Sources: [Vercel News]() Topics: [gateway](), [SDKs](), [API](), [Software]() Tags: [ai-gateway](), [api](), [evaluation](), [gateway](), [model](), [queue](), [retry](), [sdk](), [software](), [support](), [typesafe-ai]() ### AI overview Vercel's AI Gateway now supports TypeSafe AI's Jev, a probabilistic decision model that returns typed Choice, Score, and Boolean answers with probabilities. The changelog describes parallel evaluation, AI SDK 7 integration through an experimental evaluate API, example routing and review workflows, and provider options including Zero Data Retention and No Training. ### Source excerpt Jev from TypeSafe AI is now available on AI Gateway. Jev is a probabilistic decision model for software: state goes in, typed Choice, Score, and Boolean answers come out. Regular language models generate text one token at a time, which the application then parses and validates. Jev evaluates all declared questions in parallel and returns typed answers plus probabilities directly. That removes unnecessary text generation and makes it straightforward to automate clear cases while routing uncertain ones to review. TypeSafe reports Jev was up to 193.6x faster and 444.6x cheaper than LLMs on its workflow evaluations. Example use cases include: Choosing the next tool or subagent in an agent loop Deciding whether to continue, retry, ask the user, or stop Scoring urgency or risk before an action Verifying model outputs and enforcing guardrails. AI SDK 7 exposes Jev through the experimental evaluate API. Choice selects an option, Score grades an ordered rubric, and Boolean estimates the probability of true. Install the current AI SDK (AI SDK 7.0.105 onwards supports the evaluate API): Each evaluation specifies: model: the evaluation model to call, state: the shared string, object, or array to evaluate, and questions: a map of named decisions to make about that state. Call the model with typesafe-ai/jev. This example turns one support case into a queue, priority, and refund-review decision, with uncertain routing sent for manual review: The result preserves question IDs and Choice keys. TypeSafe reports separate Choice and Score confidence in result.providerMetadata.typesafe.confidence. Calibrate probabilities and confidence against labeled examples from your workflow. Jev supports Zero Data Retention and No Training, enabled per request in the example. Evaluation calls also appear in logs and custom reporting, count toward budgets, and accept other Gateway provider options in the same providerOptions.gateway object. Read the documentation on evaluation models on AI Gateway f ## Gemini Live audio DevFeed: [Gemini Live audio]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-09-15T22:47:07Z Content type: tutorial Language: en Sources: [Simon Willison's Weblog]() Topics: [speech-to-speech](), [WebSocket](), [Google AI](), [browser](), [Playback](), [implementation](), [API]() Tags: [api](), [audio](), [browser](), [gemini](), [gemini-196](), [generative-ai](), [generative-ai-1-982](), [google](), [google-416](), [llm-release](), [llm-release-231](), [llms](), [llms-1-948](), [model](), [models](), [openai](), [playback](), [release](), [speech-to-speech](), [speech-to-text](), [speech-to-text-21](), [tools](), [tools-78](), [ui](), [voice](), [websocket](), [websockets](), [websockets-21]() ### AI overview The article describes a browser-based web UI for trying Google's Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking speech-to-speech models. The implementation supports model and voice selection, an optional system prompt, voice conversations, and interruption while the model is speaking. It uses no libraries, connecting to a WebSocket endpoint and using the Web Audio API for capture and playback. ### Source excerpt Tool: Gemini Live audio Google released Gemini 3.8 Live and 3.8 Live Extended Thinking today - two new speech-to-speech models that are a similar shape to OpenAI's GPT-Live family. I pointed GPT-6 Astra Extra High at the documentation and had it build me this web UI for trying out the new models. You can select a model and voice preset, enter an optional system prompt and then start a voice conversation through your browser, including the ability to interrupt the model while it is talking. The implementation uses no libraries. It connects to the wss://generativelanguage.googleapis.com/ws/google.ai.generativelanguage.v1alpha.GenerativeService.BidiGenerateContent?key=... WebSocket endpoint and uses a Web Audio API AudioContext for both capture and playback. Here's the Gemini Live tutorial for getting started with that WebSockets API. Tags: google, tools, websockets, generative-ai, llms, gemini, llm-release, speech-to-text ## Salesforce Announces Koa Reasoning Model Built on NVIDIA Nemotron 3 Super DevFeed: [Salesforce Announces Koa Reasoning Model Built on NVIDIA Nemotron 3 Super]() Original publisher: [Read original article]() Author: Brian Caulfield Published: 2026-09-15T22:24:34Z Content type: news Language: en Sources: [NVIDIA Blog]() Topics: [Koa](), [Nemotron](), [Artificial Intelligence](), [Nvidia](), [post-training](), [ai safety]() Tags: [agentic](), [agentic-ai](), [ai](), [cloud](), [corporate](), [events](), [model](), [nemotron](), [nvidia](), [nvidia-nemo](), [post-training](), [safety](), [security](), [software]() ### AI overview At Salesforce Dreamforce, NVIDIA CEO Jensen Huang discussed AI infrastructure, safety, and enterprise adoption alongside Salesforce CEO Marc Benioff. The event coincided with the announcement of Koa, Salesforce's first CRM reasoning model, built by post-training NVIDIA Nemotron 3 Super on a proprietary synthetic dataset derived from nearly three decades of enterprise CRM deployments. ### Source excerpt Know everything. Do anything. That was the message NVIDIA founder and CEO Jensen Huang brought to Salesforce Dreamforce Tuesday, joining CEO Marc Benioff onstage in an appearance that coincided with the announcement of Koa -- Salesforce's first CRM reasoning model, built on NVIDIA Nemotron 3 Super. Huang didn't just take the stage. He walked into [...] ## 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... ## Measuring and Improving Consistency in Repeated Agent Runs DevFeed: [Measuring and Improving Consistency in Repeated Agent Runs]() Original publisher: [Read original article]() Author: Evelyn Duesterwald; Lilian Ngweta; Vatche Isahagian; Jayaram Radhakrishnan; Vinod Muthusamy; Gaodan Fang; Ashwath Vaithinathan Aravindan; Punleuk Oum; G Thomas; Merve Unuvar; Ayhan Sebin; Michał Ulewi Published: 2026-09-15T16:00:44Z Content type: article Language: en Sources: [Hugging Face - Blog]() Topics: [Benchmark](), [Ground truth / benchmark quality](), [model-deployment]() Tags: [agent](), [benchmarks](), [consistency](), [evaluation](), [gpt](), [inference](), [leaderboard](), [model](), [reports](), [standard](), [technical]() ### AI overview This article presents the Consistency Analyzer, a diagnostic for finding decision points where an agent's behavior may change across repeated runs. It introduces consistency guidelines in ALTK-Evolve and reports that they reduced the consistency gap from 24.4 percentage points to 12.0 points without reducing average accuracy. ### Source excerpt That is embarrassing onstage. In production, it is a reliability problem: a workflow that succeeded once may fail the next time a user makes the same request. For mission-critical work, such as reconciling a financial transaction or checking a contract for an obligation, that can be a showstopper. Most benchmarks hide this variability behind an average. On AppWorld, a ReAct agent using GPT-4.1 succeeded on 77.4% of runs across five repetitions. ## Microsoft drafts feel-good AI model guidelines and wants your input DevFeed: [Microsoft drafts feel-good AI model guidelines and wants your input]() Original publisher: [Read original article]() Author: Thomas Claburn Published: 2026-09-15T00:15:29Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [Artificial Intelligence](), [Microsoft](), [ai and ml]() Tags: [ai](), [ai-and-ml](), [ai-safety](), [artificial-intelligence](), [code-of-conduct](), [microsoft](), [model]() ### AI overview Microsoft has drafted aspirational guidelines for AI model behavior and is seeking public input, while acknowledging that models may still get things wrong. ### Source excerpt Redmond outlines 'aspirational' goals for model behavior, but gives itself a pass if it gets things wrong ## Understanding W8A8 INT8 LLM quantization: Accuracy and performance results DevFeed: [Understanding W8A8 INT8 LLM quantization: Accuracy and performance results]() Original publisher: [Read original article]() Author: Sana Fayyaz Published: 2026-09-14T13:01:43Z Content type: article Language: en Sources: [Red Hat](), [Red Hat Developer]() Topics: [Large Language Model](), [quantization](), [llama](), [Compression](), [Benchmark](), [dataset](), [Algorithms]() Tags: [ai-inference](), [artificial-intelligence](), [benchmark](), [benchmarks](), [compression](), [evaluation](), [hardware](), [llama](), [llm](), [model](), [models](), [performance](), [quantization](), [reasoning](), [zero-shot]() ### AI overview The article evaluates W8A8 INT8 quantization of a Llama 3.1 8B Instruct model. It describes reducing the model from 14.9 GB to 8.0 GB with SmoothQuant and GPTQ, then compares the base and compressed models on four benchmarks to assess accuracy and performance. ### Source excerpt In Understanding W8A8 INT8 LLM quantization: Half the size, better performance, same accuracy, we compressed a Llama 3.1 8B Instruct model from 14.9 GB to 8.0 GB using 8-bit integer (INT8) W8A8 quantization with SmoothQuant and Generative Pre-trained Transformer Quantization (GPTQ). The post Understanding W8A8 INT8 LLM quantization: Accuracy and performance results appeared first on Red Hat Developer. ## AI and its main promoters are not enterprise-ready, says Gartner DevFeed: [AI and its main promoters are not enterprise-ready, says Gartner]() Original publisher: [Read original article]() Author: Simon Sharwood Published: 2026-09-14T05:11:19Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [Artificial Intelligence](), [ai and ml]() Tags: [ai](), [ai-and-ml](), [enterprise](), [gartner](), [model]() ### AI overview The article reports Gartner's assessment that AI and its main promoters are not ready for enterprise use, citing the rapid pace of model-makers and their lack of concern when changes break things. ### Source excerpt Model-makers move too fast and don't care when they break things ## A study of sequence weighting at scale DevFeed: [A study of sequence weighting at scale]() Original publisher: [Read original article]() Author: Alex Renda Published: 2026-09-14T00:00:00Z Content type: article Language: en Sources: [Jane Street]() Topics: [scaling laws](), [datasets](), [Deep neural networks](), [Chinchilla scaling law]() Tags: [data](), [machine-learning](), [model](), [scaling-laws](), [training]() ### AI overview The article studies how sequence weighting affects language-model training across model scales. It reports non-monotonic behavior: small-to-medium models increasingly learn data-specific patterns in proportion to sequence weights, while large models can learn all patterns in the data more independently of those weights. ### Source excerpt TL;DR: We study the scaling laws of data weighting across in-house and open-weight LMs, finding non-monotonic behavior across scales. We vary the weight assigned to sequences during training and measure how strongly the model's loss reduction on a sequence depends on the sequence's weight. Taken together, our results are consistent with a general trend: as models transition from small to medium scale, they transition from learning general patterns independent of data weight to learning data-specific patterns proportional to the data weights. As models then transition from medium to large scale they are able to learn all patterns present in the data, once again independent of data weight. ## Interpreting Pangram DevFeed: [Interpreting Pangram]() Original publisher: [Read original article]() Author: Armin Ronacher Published: 2026-09-14T00:00:00Z Content type: opinion Language: en Sources: [Armin Ronacher]() Topics: [Artificial Intelligence](), [Large Language Model](), [datasets](), [Training AI Models]() Tags: [ai](), [llm](), [model](), [openai](), [training](), [training-data]() ### AI overview The article discusses Pangram, an AI-text detector that classifies passages as human-written, AI-generated, or mixed. It explains that Pangram manufactures training data from human-authored text and LLM-generated rewrites and edits, then describes an experiment using Opus 5 to generate text intended to read as entirely AI-generated. ### Source excerpt Yesterday David Sacks wrote a tweet and within a few minutes people did, what they usually do, and they asked Pangram if it was AI. And Pangram said it's entirely AI generated. To which David replied that these AI detectors are bogus. Now Pangram has a pretty low false positive rate, but if you have ever used an LLM as a writing assitant, you will have probably noticed that it claims your posts 100% AI, even though you don't feel like they are. Pangram itself is a trained model, that attempts to detect segments of text as being definitely human, definitely AI and a mixture of the two. If you want to know how it works, they published a paper. The short summary is that they are manufacturing its own training data by starting from collections of known human authored text. An LLM is then tasked to understand the text and write a fresh new text on the same topic. They also let the LLM perform partial edits on that original human text and through that they can pick up on these co-authored details. Pangram claims their model to have rates of 0.0041% false AI accusations and 0.34% missed AI text. So now that we know this I figured it might be fun to have an LLM re-create David's tweet. I first came up with a prompt. And when I say I came up with that prompt I in fact used an LLM to propose to me from that tweet what I might want to say for the structure. I'm sure if you ask Pangram about if the above text is AI, it will probably say so, but that's not really the point. The point is that I then used Opus 5 to generate a text which reads entirely AI generated. If you are curious, this is the prompt I used: Original prompt to generate a David Sacks like tweet Please write an 8 paragraph tweet about "Pacing the Frontier", not longer than 350 words. Context of this tweet is this post by Dario: https://darioamodei.com/post/we-must-pace-the-frontier And this tweet by Sam Altman: "I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we' ## "Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason DevFeed: ["Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason]() Original publisher: [Read original article]() Author: Adrian Bridgwater Published: 2026-09-13T14:21:46Z Content type: news Language: en Sources: [The New Stack]() Topics: [Mixture of Experts (MoE)](), [cohere](), [Artificial Intelligence](), [Benchmark](), [Inference](), [qwen](), [gemma4](), [Google]() Tags: [ai](), [ai-infrastructure](), [ai-models](), [aya](), [benchmark](), [cohere](), [gemma](), [google](), [inference](), [large-language-models](), [mixture-of-experts-moe](), [model](), [open](), [qwen]() ### AI overview Cohere's North Small Translate is an open-weight mixture-of-experts machine translation model covering 50 languages. The article discusses its non-reasoning design, sovereign AI positioning, deployment options, efficiency claims, and reported WMT26 benchmark comparisons. ### Source excerpt Enterprise AI company Cohere announced North Small Translate last week, a mixture-of-experts (MOE) open-weight machine translation model that works across The post "Machine translation is still broken for most of the world's languages": Cohere builds non-reasoning for a reason appeared first on The New Stack. ## Domain-Driven Design Clearly Explained DevFeed: [Domain-Driven Design Clearly Explained]() Original publisher: [Read original article]() Author: Nikki Siapno Published: 2026-09-12T12:52:28Z Content type: tutorial Language: en Sources: [Level Up Coding System Design Newsletter]() Topics: [Domain-driven design (DDD)](), [Code](), [systems]() Tags: [concepts](), [developers](), [domain](), [domain-driven-design](), [engineering](), [model](), [patterns]() ### AI overview An introduction to Domain-Driven Design (DDD), explaining how shared business language and bounded contexts help teams model complex products and keep code aligned with the domain. ### Source excerpt The mental model that makes DDD click. ## Teaching AI to Reason Through Detection Triage DevFeed: [Teaching AI to Reason Through Detection Triage]() Original publisher: [Read original article]() Author: Amol Khanna - Manu Nandan - Cristian Viorel Popa - Joan Pujol-Roig - Diana Bolocan - Laura Vasilie - Alexandru Apostu - Chase Helwig - Mihaela Gaman - Mickey Brautbar - Edward Raff - Chase Midler - Sv Published: 2026-09-12T11:17:51.295154Z Content type: article Language: en Sources: [Blog]() Topics: [AI Chat]() Tags: [agentic](), [agentic-soc](), [ai](), [automation](), [chain-of-thought](), [classification](), [cybersecurity](), [language-models](), [llm](), [model](), [nemotron](), [nvidia](), [open](), [reasoning](), [research](), [security](), [soc](), [windows]() ### AI overview CrowdStrike describes research on a reasoning-enabled language-model classifier for security detection triage. The model produces a verdict and an auditable rationale, with the stated goals of improving accuracy, transparency, and safe alert automation. ### Source excerpt New CrowdStrike research shows how step-by-step reasoning can improve detection triage accuracy, transparency, and safe automation. ## HeyGen x Google Cloud: Bringing Avatar IV to TPUs DevFeed: [HeyGen x Google Cloud: Bringing Avatar IV to TPUs]() Original publisher: [Read original article]() Published: 2026-09-12T11:04:33.891311Z Content type: article Language: en Sources: [Google Developers Blog]() Topics: [Google](), [Cloud](), [Artificial Intelligence](), [Optimization](), [AI Infrastructure](), [Transformer](), [PyTorch](), [Compiler](), [Streaming](), [real-time](), [Hardware](), [API]() Tags: [ai](), [ai-infrastructure](), [api](), [cloud](), [code](), [compiler](), [engineering](), [generation](), [google](), [hardware](), [model](), [optimization](), [performance](), [pytorch](), [real-time](), [streaming](), [time](), [tpu](), [transformers](), [video]() ### AI overview HeyGen and Google Cloud describe porting the 18B+ parameter Avatar IV talking-head video generation pipeline to an eight-chip Trillium TPU host. Using torchax, JAX, XLA, FSDP sharding, Ulysses sequence parallelism, and custom Pallas kernels, the team improved performance by 1.86x for real-time chunked streaming while preserving output quality through strict quality gates. ### Source excerpt HeyGen ported their 18B+ parameter Avatar IV video generation model to Google Cloud's Trillium (v6e) TPUs via torchax and XLA, utilizing FSDP and Ulysses sequence parallelism across an eight-chip mesh. To achieve a 1.86x speedup for real-time streaming, the engineering team pipelined exposed all-to-all collectives, aligned sparse attention block sizes to eliminate mask padding, and bypassed softmax serial dependencies using a precomputed Cauchy-Schwarz upper bound. These custom Pallas kernel and compiler optimizations were deployed only after passing rigorous two-tier quality gates to guarantee byte-identical or mathematically equivalent pixel outputs. ## How to use Google microbenchmarks for evaluating TPU performance DevFeed: [How to use Google microbenchmarks for evaluating TPU performance]() Original publisher: [Read original article]() Author: Junjie Qian; Chi Shuen Lee; Yu-Hsuan (Amy) Lin; Haixiong (Sean) Wang Published: 2026-09-12T11:04:33.891311Z Content type: tutorial Language: en Sources: [Google Developers Blog]() Topics: [Google AI](), [Inference Performance]() Tags: [architecture](), [benchmarks](), [compute](), [developers](), [google](), [guides](), [hardware](), [how-to](), [kernel](), [machine-learning](), [memory](), [mesh](), [metrics](), [model](), [open](), [open-source](), [performance](), [scale](), [software](), [tpu]() ### AI overview A tutorial on using Google's TPU microbenchmark suite to measure network, compute, memory, host-transfer, and attention performance. The results can establish a Roofline baseline and guide workload-specific optimization. ### Source excerpt Google's open-source TPU microbenchmark suite provides developers with granular performance metrics across Network, Compute, HBM, Host Transfer, and Attention components to validate real-world hardware capabilities. By leveraging these benchmarks to establish a Roofline model, engineers can accurately diagnose whether their machine learning workloads are compute-, memory-, or network-bound. This empirical baseline directly guides targeted software optimizations--such as kernel tuning, mesh sharding, and rematerialization--to maximize hardware utilization for large-scale model deployments. ## Agent and Model Evaluations in Gemini Enterprise Agent Platform are now GA DevFeed: [Agent and Model Evaluations in Gemini Enterprise Agent Platform are now GA]() Original publisher: [Read original article]() Author: Alex Martin; Dima Melnyk Published: 2026-09-12T11:04:33.891311Z Content type: release Language: en Sources: [Google Developers Blog]() Topics: [ai observability](), [SDKs](), [ci]() Tags: [agent](), [agents](), [ci](), [cli](), [development](), [enterprise](), [evaluation](), [experiments](), [gemini](), [llm](), [metrics](), [model](), [platform](), [production](), [telemetry](), [testing](), [traces]() ### AI overview Gemini Enterprise Agent Platform's evaluation service is generally available. It provides consistent evaluation of agents and models across local experiments and production traffic, with pre-built metrics, adaptive rubrics, custom metrics, simulators, and workflow integrations. ### Source excerpt Agent Platform's evaluation service is now generally available, providing developers with a unified engine to measure agent quality consistently across local development experiments and live production traffic. You can evaluate agents using over 20 pre-built metrics, DeepMind-backed adaptive rubrics, or custom code-based and LLM-as-a-judge metrics stored in a centralized, versioned registry. The service integrates directly into existing workflows via the Agent Platform SDK, agents-cli, and ADK, offering built-in user and environment simulators to automate complex multi-turn testing and streamline CI pipelines. ## OpenRouter provider fallbacks can cause inconsistent model behavior DevFeed: [OpenRouter provider fallbacks can cause inconsistent model behavior]() Original publisher: [Read original article]() Author: Simon Willison Published: 2026-09-11T22:49:18Z Content type: opinion Language: en Sources: [Simon Willison's Weblog]() Topics: [API](), [Large Language Model](), [Software]() Tags: [ai](), [ai-2-236](), [api](), [cost](), [generative-ai](), [generative-ai-1-982](), [llms](), [llms-1-948](), [model](), [models](), [news](), [openrouter](), [openrouter-32](), [providers](), [reasoning](), [serve]() ### AI overview OpenRouter's automatic provider fallbacks can produce inconsistent behavior for the same model because providers use different serving software, optimizations, and settings. The article notes differences in vision support and reasoning-effort handling, and points to provider.only and /endpoints for controlling or inspecting routing. ### Source excerpt So you want to use OpenRouter? One of OpenRouter's selling points is that it "handles fallbacks automatically and picks the most cost-effective option for each request", so you can call a single API endpoint for a model and get routed to the best available backend provider. Mohamed Moustafa points out a whole set of ways that this can cause you problems. Different providers run different serving software with different optimizations and settings, which means that the same OpenRouter endpoint can serve model requests that behave in different ways. Some providers even lack vision capability for vision models, and the way the reasoning effort option is processed can differ as well. Thankfully you can control which provider is routed to using the provider.only option. The /endpoints method returns the list of available providers for a specific model ID. Via Hacker News Tags: ai, generative-ai, llms, openrouter ## Cohere's new translation model is open weights -- but not for commercial use DevFeed: [Cohere's new translation model is open weights -- but not for commercial use]() Original publisher: [Read original article]() Author: Meredith Shubel Published: 2026-09-11T17:50:11Z Content type: news Language: en Sources: [The New Stack]() Topics: [Inference](), [moe]() Tags: [ai](), [ai-infrastructure](), [ai-models](), [ai-strategy](), [api](), [cohere](), [hugging-face](), [inference](), [mixture-of-experts](), [model](), [models](), [open](), [production](), [release]() ### AI overview Cohere released North Small Translate 1.0 as open weights under CC BY-NC 4.0, allowing download, evaluation, and study but requiring a commercial agreement for production use. Commercial deployment requires a license and use of Cohere's managed Model Vault platform. ### Source excerpt This week, Cohere released North Small Translate 1.0 under a CC BY-NC 4.0 license: the weights are there to download, The post Cohere's new translation model is open weights -- but not for commercial use appeared first on The New Stack. ## DeepSeek's new model sets a template for powerful LLMs that run lean DevFeed: [DeepSeek's new model sets a template for powerful LLMs that run lean]() Original publisher: [Read original article]() Author: Tobias Mann Published: 2026-09-11T07:15:00Z Content type: news Language: en Sources: [www.theregister.com - Articles]() Topics: [deepseek](), [Large Language Model](), [Caching](), [Latency](), [Google]() Tags: [ai-and-ml](), [cache](), [compute](), [cost](), [datacenter](), [deepseek](), [embedding](), [flash](), [google](), [gpu](), [latency](), [llm](), [llms](), [model](), [release]() ### AI overview DeepSeek V4.1 Flash is a larger LLM whose architectural changes aim to reduce serving memory and compute needs. The article highlights lower KV-cache consumption, improved prompt processing, and an N-gram-based conditional memory module. ### Source excerpt DeepSeek V4.1 Flash proves that just because you build a bigger model doesn't mean you need more GPUs to serve it [Next page]()