model
Thinking Machines Lab's 952B-parameter open model, an American exception to the sub-130B ceiling.
Verbalization training made three models voice test suspicion 2.4 to 2.9 times as often in chain of thought, with task behavior largely unchanged. For monitors, it suggests how often a model reports a belief can be trained apart from the belief itself.
Reality
Bloomberg reports that US frontier labs and the US government are increasingly worried about distillation attacks, and that foreign actors are buying logs of conversations held on legitimate accounts. The detection Anthropic describes happens at request time.
Reality
The Information now puts the raise at $5bn to $6bn on at least $40bn pre-money, with Nvidia writing roughly half the cheque, which makes this a compute-supply arrangement for anyone buying Tinker cluster time.
Reality
Gemini 3.8 Flash is the third Flash model in six weeks, and the scores that came with it are self-reported. Commenters on the Ars Technica thread landed on the dull fix first: keep your own eval set.
Reality
The lab named tamper resistance an open question 40 days after Inkling's weights shipped. The award is platform compute, priced by the grantor and spendable only on the approved study.
Reality
A Fortune columnist reads the 2026 AI economy as an unstable system of frontier labs, Chinese open weights and app companies. The checkable part is price, not payoff.
Reality
Hugging Face now hosts 2.96 million public model repositories. The ones anyone actually pulls number in the tens of thousands, and the monthly parameter ceiling has been set in China all year.
Reality