Open source

AINVIDIA publishes a GR00T N2 checkpoint, step 18200, as a native split ONNX export on Hugging Face for SSD pickup and placement. The model trained on 196 pickup and 149 placement episodes, with 310 training and 35 validation episodes, and one shared set of graphs serves both tasks through a host-side task selector. The package is not a TensorRT engine or a robot-ready policy, and NVIDIA does not assert full ONNX/eager numerical parity or robot success rates.

AIAi2 says its Olmo Hybrid paper shows a model combining transformer attention with linear recurrent layers reached the same MMLU accuracy as Olmo 3 7B using 49% fewer training tokens. The lab says a next Olmo model with a hybrid mixture-of-experts architecture is in pre-training, and that it released Olmo-core 3 for training large mixture-of-experts models. Ai2 also released AstaBrief, an open-weights model that generates cited reports from research questions and retrieved literature.

AIPrime Intellect says it rewrote its Prime Agent coding tool in Rust, using a swarm of more than 2,000 agents over two weeks. The company reports cold start to typing about 13 times faster than the TypeScript version, and memory use over 80% lower after startup. Prime Agent remains open source and adds native Windows support in beta and Homebrew installation.

Why it matters: The post shows how a multi-agent swarm rewrote a coding agent with parity checks, giving a concrete case of agent-driven software engineering with measured results.

AIPrime Intellect says Prime Agent is now open source and that its Rust rewrite is available today. The full blog post covers verification loops, multi-agent orchestration, and performance experiments.

AIPrime Intellect says users can update their Prime Agent by running the command prime-agent update. The post links to the project's GitHub repository, PrimeIntellect-ai/prime-agent.

AISierra has published a draft of the Personal Agent Protocol, known as Poppy, and named 35 additional design partners, including Adyen, Bank of America, Mastercard, OpenAI, PayPal, and Visa. Under the protocol, companies publish a /.well-known/poppy.json discovery file, and personal agents start sessions, identify themselves, and sign in through OAuth with session tokens limited to approved access. The company says the draft will be followed by design workshops and a reference implementation over the next month.

Why it matters: The draft specifies how personal agents identify themselves, obtain customer-approved access, and work with company websites, APIs, or agents, which helps readers assess its practical effect on agent-driven transactions.

AIAlibaba's Qwen team released Qwen-Image-2.1-Turbo, an accelerated checkpoint of Qwen-Image-2.1 that generates and edits images in 8 denoising steps instead of 40. The model keeps the same 7B architecture and offers a hosted API at CNY 0.1 per image, while its weights are under a Qwen Research License that requires separate permission for commercial self-hosting.

AIQwen has released its Turbo image model with open weights, giving image app builders an option to run it in-house instead of relying on hosted API pricing. The post does not specify model size, benchmarks, or pricing.

AICloudflare has published a new model called clef-omni on Hugging Face, according to a post from Julien Chaumond, who owns the account. The post links to the model page but gives no further details about its size, capabilities, or benchmarks.

AIX launches an X API Engineer in Grok Bot that suggests project ideas, helps build and test them, and deploys them for others to use. The post, from the Cursor account's owner, invites developers to build on the X API with this tool.

AIGoodfire announces a partnership with Baseten to bring real-time safety monitoring to open models. The companies are launching Project Beacon, which combines inference with in-line monitoring so safety signals are caught proactively.

AIMicrosoft says the full Word, Excel, and PowerPoint apps are now rolling out to the Copilot app in Frontier. Users can create, edit, and collaborate on documents in one connected workspace, with branded templates and version history. The workspace also offers a library of frontier models, with an Auto router that picks a model for each task.

AIOpenAI's developer account announces composer predictions in Codex, which suggest a user's next message based on the conversation and how the user writes. During the beta, the feature is included at no additional cost for eligible Pro users.

AICloudflare releases Clef-omni, an open-weight decision model that accepts audio, video, image, and text input in a single API call. Clef-flash's price falls from $0.09 to $0.038 per M input tokens, while its hosted context window drops from 64k to 24k. Cloudflare also reports median latency reductions of 1.7 to 2.0 times for the Clef model on Workers AI.

AICline is offering Solar Mini 4 free, a new 35B mixture-of-experts model from Korean lab Upstage with 3B active parameters. It has a 524K context window and runs at 208 tokens per second. Cline says it scores 24 on the AAII, the highest of any model at 3B active and within one point of Nemotron 3 Ultra, which uses 55B active.

AISuno says the wavetable synth, musical typing keyboard, and expanded plugin views in Suno Studio can now float anywhere on screen, including a second monitor. Studio is included with Premier.

AIAnthropic has launched Cyber Mission, a long-term program to protect critical infrastructure and open-source software from cyberattacks. A free OSS AI scanner will regularly check open-source projects, flag and explain vulnerabilities, and suggest patches. Anthropic expects over 90 percent accuracy, but reports ship without human review and may contain errors.

AIBaseten announces Project Beacon with Goodfire AI, adding inline safety controls to model inference. Goodfire's activation-based monitors read a model's internal activations during generation, so policies can flag unsafe events before output reaches a user or tool. Baseten plans to release the capabilities over the next several months with selected models and early partners.

AIMistral releases Voxtral Mini 4B Realtime Arabic, a streaming speech-to-text model for Arabic dialects and Modern Standard Arabic under the Apache 2.0 License. The model has about 4.4 billion parameters, is fine-tuned from Voxtral-Mini-4B-Realtime-2602, and reaches an average 8.82% Character Error Rate across seven Arabic benchmarks at a 480 ms transcription delay. It can be run with vLLM or Transformers 5.2.0 or later.

AIMistral AI has released LIDstral-Arabic, a fast classifier that identifies Modern Standard Arabic, Arabic dialects, and non-Arabic languages written in Arabic script across 51 classes. On Moroccan Darija, it scores 88.67% F1, versus 72.65% for LahjatBERT ALDi CL and 71.28% for GlotLID v3, across 84,870 evaluation examples. The model runs on CPU and is available from a private Hugging Face repository under Apache 2.0.

AIAmazon Bedrock Managed Agents, powered by OpenAI, entered public preview, and OpenAI's GPT-6 Astra, GPT-6.1 Sol, and GPT-6.1 Luna became generally available on Amazon Bedrock. AWS also released Strands Decider 2B, a 2B-parameter open source decision model that answers in about 115ms locally, and said the Strands harness uses 28 percent fewer tokens than popular harnesses while matching their accuracy.

AIQwen has published Qwen-Image-2.1-Turbo on Hugging Face, an accelerated checkpoint of Qwen-Image-2.1 for text-to-image generation and image editing with 8 denoising steps. The checkpoint uses the same 7B visual generation architecture, loads directly with QwenImage21Pipeline in Diffusers, and includes its recommended sampling schedule. It defaults to CFG=1 and uses prefix KV caching to reuse text and reference-image context across steps.

AIGoogle AI announces that SynthID.com is now available globally in English for verifying AI-generated images, video, and audio. The post also lists Nano Banana 2.1, EmbeddingGemma 2, Gemma 4 with BOTANIC-1, a Gemini business agent, and Guided Vision in Gemini Live.

AIAlibaba's Qwen team says Qwen-Image-2.1-Turbo is an accelerated checkpoint of Qwen-Image-2.1 on the same 7B architecture, now with open weights. It generates 2K images from text in 8 denoising steps and supports natural-language edits, with Pro and Turbo APIs also live.

AITencent has open-sourced Youtu-Parsing-Omni, a 5B-parameter omni-modal model that outputs a single structured JSON covering layout, text, tables, formulas, ASR, OCR, and video segments. It scores 96.96 Overall on OmniDocBench, the highest among the compared models, and ships with weights on Hugging Face, a vLLM plugin, and inference examples.

AINVIDIA has released a deployment model for Agile One S SSD pickup, based on GR00T N1.7 checkpoint 58000 and using three cameras: ego, left wrist, and right wrist. The repository republishes ONNX graphs, external tensor files, and two existing TensorRT BF16 engines without retraining or re-export, and the original export reported a numerical warning that full FP32, node, and BF16 parity did not pass all tolerances. The files are not a certified robot deployment or safety qualification.

AINVIDIA published the Agile One S Walk GR00T N2 checkpoint 1680, a walking deployment model with four cameras, on Hugging Face. The repository includes ONNX graphs, TensorRT BF16 plans/engines, and the original checkpoint files, republished without retraining or re-export. The shared Cosmos-Reason1-7B dependency and the Isaac/GR00T runtime must be set up separately, and the files are not a robot safety qualification.

AIJetBrains released Mellum2.1, a 12B mixture-of-experts coding model with 2.5B active parameters under Apache 2.0, emphasizing agentic programming. Under high load, its inference throughput in tokens is nearly twice that of Qwen3.5-9B in JetBrains' comparison, and multi-token prediction (MTP) speeds single-request responses by about 1.6x. The model is available on Hugging Face for local or private-infrastructure deployment, with GGUF and vLLM MTP support announced for later.

AIHuawei-backed openJiuwen has open-sourced AgentOS for Enterprise under Apache 2.0 on GitHub and AtomGit, targeting multi-agent coordination, memory-based self-evolution, multi-tenant isolation and fault recovery. Huawei Connect 2026 also introduced an all-in-one appliance built on it, which the launch information says enables an end-to-end private deployment in hours.

AIMarkTechPost publishes a hands-on tutorial implementing RRSI (Regularized Recursive Self-Improvement), a method that lets an LLM agent revise its own harness around a frozen model. The full loop drafts edits with Claude Opus on Vertex AI and scores them in Docker benchmarks, but the edit-selection rules are plain Python that the tutorial runs in a simulated environment with a calibrated noise band.