Ramine Roane says AMD is currently the leading company contributing code to vLLM, with 502 core commits in 90 days, 13% of all organization contributions, and nearly four times Nvidia. He also credited Red Hat, IBM, Embedded LLM, and Inferact.
WIRED reports that millions of people have downloaded Meta’s AI agent Muse, and that using it comes with privacy costs: Muse creates detailed profiles of users’ friends and family.
Cloudflare AI Search is generally available, and hybrid search is enabled by default. Workers AI embeddings are included in monthly usage and do not appear on the Workers AI bill.
Aravind Srinivas says Perplexity is starting to deploy Perplexity Computer on Vera, which he calls far better than x86. The stated goal is to own the sandboxes and optimize them for that silicon.
Aleph Alpha’s Kolibri has 78 billion parameters, with 3.46 billion active, and a context window of up to 1 million tokens. The weights are available to run on local hardware under Apache 2.0.
Patrick Erichsen says ClawHub runs two open-source security scanners on every upload and a judge harness over both results. He says the team worked with Tencent Hunyuan and NVIDIA AI on ClawScan, and he believes this launch makes ClawHub the most secure place to find and install agent skills.
Prime Intellect has introduced Prime Inference and says it has served trillions of tokens for reinforcement learning and dedicated customer deployments.
OpenRouter has released Model Router Benchmarks to compare seven model routers side by side across quality, speed, and cost on six benchmarks, covering routers such as Jev Router, Unbiased Pareto, and NVIDIA Switchyard.
Three models are now available on Ollama for /v1/systemone: Nimble, Tev1 4b, and Tev1 0.8b, each with an Ollama library page.
Ollama says Jev-like decision models now run entirely on a local machine. Nimble can be installed with ollama pull nimble and used through a new local /v1/systemone API.
Artificial Analysis has open-sourced AA-AgentPerf-Local, which replays recorded agent sessions to measure local inference speed, and published initial 4-bit results for four systems. On models that fit in 32 GB, the GeForce RTX 5090 finished more than 3.5 times faster than the other systems tested.
Zixuan Li says GLM-5.3 has helped defend 389 open-source projects and found 4,249 potential vulnerabilities so far. OpenVuln remains free, and findings go privately to maintainers.
Artificial Analysis says that one year ago GPT-5 mini (high) led its Intelligence Index at 17 on v4.3 and about $0.05 per task. As of last week, Claude Opus 5.5 holds the highest score to date at 58 and about $5.98 per task.
Polymarket reports that Robinhood has launched AI agents that can trade stocks for users, monitor markets overnight, and automatically execute trades when specified conditions are met.
Ling 3.1 Flash is now free in Command Code on every plan and subscription. Command Code lists the model at 560 billion total parameters and 25 billion active parameters.
Manus is expanding the inference providers users can connect through Manus Flex. OpenRouter, FireworksAI_HQ, and modal are the initial partners, and OpenRouter says users add their key in Manus Flex to run agents on a chosen model.
Codex CLI 0.159.2 is out. It stops console windows from flashing on Windows, makes background process launches quieter, and cleans up sandboxed command runs.
Shareable profiles bring Sites and plugins together so others can find and reuse them. They are available to Free, Go, Plus, Pro, and Business users, with Enterprise, Edu, and Healthcare plans coming soon. Shared skills stay inside the workspace.
NVIDIA AI says organizations can use NVIDIA OpenShell as an open source option for governing agents with OpenClaw Enterprise.
v0 now offers GPT-6.1 Sol. Vercel Developers also says the model is on AI Gateway as openai/gpt-6.1-sol and improves on GPT-6 Sol for coding, computer use, multi-step workflows, and complex document analysis.