Mistral has released Timestral, a family of time-series foundation models ranging from 5 million to 209 million parameters. The largest leads the GIFT-Eval, TIME and fev-bench benchmarks among zero-shot models and runs on a CPU.
The World Bank is urging developing countries to start building 'small AI' models that run on mobile phones with weak connectivity. Its chief knowledge officer, Paschal Donohoe, said AI could speed up development in emerging economies, provided it is adapted to local needs.
Kyutai, working with Inria and École des Ponts, has released REKI, a 2.4-billion-parameter vision-language-action model for robotic manipulation, trained solely on public data and equipped with a low-cost memory of recent events. On three out-of-distribution benchmarks, only models roughly twice its size outperform it, and its weights are publicly available.
Google cofounder Sergey Brin said in a roundtable on Gemini 4 Argon that the company is starting to get onto a path of recursive self-improvement, with researchers and models working together. He cited the Flash models released before Gemini 4 as evidence.
Interconnects analyzes how Meta's latest moves reshape both open and closed frontier AI. The newsletter's verdict is that US open-weight models are back.
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.