Breaking
POLICY Claude Fable 5 and Mythos 5 Suspended Worldwide After US Export Order /// MODEL OpenAI ships GPT-5.5 Turbo with a 1M-token window /// RELEASE Google opens Gemini 3 Ultra preview to developers /// POLICY EU AI Act enforcement begins — what builders must do now /// FUNDING Anthropic closes a record raise as the model wars escalate /// RESEARCH An open-weights model just topped a closed frontier benchmark /// RELEASE xAI launches Grok 5 with real-time vision /// POLICY Claude Fable 5 and Mythos 5 Suspended Worldwide After US Export Order /// MODEL OpenAI ships GPT-5.5 Turbo with a 1M-token window /// RELEASE Google opens Gemini 3 Ultra preview to developers /// POLICY EU AI Act enforcement begins — what builders must do now /// FUNDING Anthropic closes a record raise as the model wars escalate /// RESEARCH An open-weights model just topped a closed frontier benchmark /// RELEASE xAI launches Grok 5 with real-time vision
● ON AIR
StepTenNews// the freshness desk
we file within seconds of the wire← stepten.io
Breaking · research

An open-weights model just topped a closed frontier benchmark

The gap narrows again. We dig into the eval, the caveats, and whether it holds up in real agent work.