OpenAI's upcoming 'Astra' model family enables multi-agent collaboration over days, debuting by solving ten previously open math problems.
DeepSeek-V4-Flash-0731 is a 304B parameter model optimized for agentic workflows, punching well above its weight.
Anthropic's official policy statement outlines their stance on open-weights, balancing safety risks against the benefits of open-source innovation.
ByteDance's Seedance 2.5 generates unified 30-second video and audio clips, tripling the duration of Google's Gemini Omni Flash.
Supabase open-sourced an Apache-2.0 benchmark to evaluate coding agents on real-world database tasks, schema building, and RLS policies.
A field report warns that coding agents accelerate scientific software modernization by 60x but produce highly convincing, silent errors.
Google introduced Managed Agents for Gemini 3.6 Flash, adding native hooks and triggers to simplify agentic workflows.
OpenAI has reduced the Codex model context size from 372k to 272k tokens, impacting long-context retrieval strategies.
Simon Willison co-released 'smevals', a lightweight evaluation suite designed to test models, prompts, and agent harnesses.
Google DeepMind launched Lyria 3.5 in Flow Music, delivering significant upgrades to lyric generation, vocals, and musicality.
OpenAI is granting 100,000 academic researchers free access to its advanced models to accelerate scientific computing and discovery.
Google announced Gemini Robotics ER 2, powering two new physical robot platforms named Duo and Apollo.