Official updates
Announcements from official sources, ordered by publication time.
14 updates · Last 30 days
October 19, 2026 · UTC
Introducing Clairic, a new System One class of extraction models
Cloudflare · Official · Models · Published
AI summary: Cloudflare introduces Clairic, a preview System One extraction model that returns typed values, estimates, and forecasts from text or images in a single pass without generating output tokens.
Official article · Also seen on Hacker News
TorchNeuron: running HF models on AWS Trainium is easier than ever
Hugging Face · Official · Models · Published
AI summary: TorchNeuron adds a native PyTorch backend for AWS Trainium, letting Hugging Face libraries run models with familiar PyTorch APIs for inference, training, and distributed execution.
October 16, 2026 · UTC
Introducing Reflection Beam: an open-weight model for agentic coding on AWS
Amazon Web Services · Official · Models · Published
AI summary: This AWS post introduces Reflection Beam, an open-weight model for agentic coding, and documents deployment on SageMaker JumpStart or HyperPod Inference with JumpStart, SGLang, or vLLM.
Welcome Beam: a 501B open model you can own, customize, and scale
Hugging Face · Official · Models · Published
AI summary: The Hugging Face announcement introduces Beam, a 501B-parameter open Apache 2.0 MoE model designed for long-context reasoning, coding, and agentic workloads.
October 15, 2026 · UTC
NVIDIA Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning
NVIDIA · Official · Models · Published
AI summary: NVIDIA presents Nemotron 3.5 Super VL, an open 120B-parameter multimodal MoE model designed for customizable enterprise agents requiring reasoning, tool use, and text and visual understanding.
Open From Data Center to Desktop: NVIDIA Expands Nemotron Family With Highest Efficiency Super VL and Bolt Models
NVIDIA · Official · Models · Published
AI summary: NVIDIA introduced three open Nemotron 3.5 models spanning enterprise servers to edge devices, highlighting Super VL’s multimodal agent performance and Bolt models’ local deployment efficiency.
Welcome Muse Spark Open: open-weight frontier multimodal mixture-of-experts model from Meta
Hugging Face · Official · Models · Published
AI summary: The page introduces Meta’s Muse Spark Open, an open-weight multimodal mixture-of-experts model for agentic coding, long-context reasoning, tool calling, and text generation from text, images, audio, and video.
October 14, 2026 · UTC
Amp puts AI agents to work beyond code with OpenAI models
OpenAI · Official · Models · Published
See how Amp pushes the frontier of what agents can do with GPT-6 Astra, Ultrafast, GPT-Live-1, and Sign in with ChatGPT.
October 13, 2026 · UTC
Developing Gemini 4 Argon responsibly and safely
Google · Official · Models · Published
AI summary: Google describes Gemini 4 Argon’s safety approach, reporting stronger cyber and biology safeguards, real-time monitoring, reduced shortcut behavior, and 0.7% attack success on an indirect prompt-injection benchmark.
Gemini 4 Argon for Robotics
Google · Official · Models · Published
AI summary: Gemini 4 Argon is presented as a general-purpose physical reasoning model for robotics, scoring 87% on ERQA and supporting control, simulation reconstruction, physical understanding, and self-improvement workflows.
Gemini 4 Argon: Built to solve your hardest problems
Google · Official · Models · Published
AI summary: The page introduces Gemini 4 Argon, a frontier model for complex coding, professional, scientific, cybersecurity, and robotics tasks, claiming leading benchmark performance and broad real-world impact.
Gemini 4 Argon: Built to solve your hardest problems
Google · Official · Models · Published
AI summary: Google introduces Gemini 4 Argon, a globally available frontier model for complex coding, business, scientific, cybersecurity, and robotics tasks, claiming leading benchmark performance and substantial real-world productivity gains.
October 12, 2026 · UTC
ReplaySSM: Rethinking State Storage for Faster Hybrid-Model Decoding
NVIDIA · Official · Models · Published
Hybrid models interleave state-space model (SSM) layers with a few attention layers. NVIDIA Nemotron 3, Qwen3.5 through Qwen3.8, Kimi-K3, and GLM-5.3-Flash are...
Run multi-node inference workloads using Ray LLM Deep Learning Containers
Amazon Web Services · Official · Models · Published
Learn how to deploy one large language model across two GPU nodes on Amazon EKS using the multi-node Ray LLM Deep Learning Containers, with separate prefill and decode nodes behind one OpenAI-compatible endpoint. The pre-built container ships the full inference stack, so you configure the deployment instead of assembling it.