StepFun has published Step 5 Preview as an open-weight model, with FP8 and BF16 checkpoints on Hugging Face and ModelScope. The sparse mixture-of-experts model is built for long agentic tasks, with 600 billion parameters in total, 27 billion active per token and a context of up to one million tokens.
TypeSafe launched Jev, a decision model that returns a label, a score or a yes/no probability from answers the user defines, on 15 September. In roughly four weeks at least 20 rivals appeared, from Cloudflare, OpenAI, Perplexity, Liquid AI, AWS's Strands Labs and independent developers.
DeepSeek has released V4.1-Flash-1017, a new checkpoint of V4.1-Flash available on its website, app and API. The company says it scores higher, finishes tasks faster and writes shorter outputs than the September release, calling it a gain on both performance and cost, and it is also opening a search API.
Microsoft's AI research group has detailed MAI-Base-2, which it says is significantly more efficient than leading open-source alternatives. It is a base model rather than a finished assistant, meant to be turned into a full LLM through further training.
Vercel's AI Gateway now offers DeepSeek V4.1 Flash 1017, a minor update to V4.1 Flash. It is described as better at reasoning, coding, agentic tasks and image understanding, while finishing tasks faster with shorter outputs, which should make agent runs cheaper.
Security firm XBOW won $250,000 from Google's Chrome bug bounty programme with a full exploit chain built using GPT-5.6 Sol. One of its researchers said such work once took a team of four six to nine months, while the awarded submission took six days.