Models & Research

AI Plus subscribers paying $4.99 lose Pro as well. In exchange, the Pro tier picks up Deep Think, which used to start at $99.99 a month.

Since October 1, Google has stopped taking product vulnerability reports. The announcement puts the restart in the first quarter of 2027.

To build the index, Claude scored about 19,000 O*NET task descriptions and hunted for a real robot behind each one. A side finding from 50 years of backtesting: robots gain roughly 2% of previously impossible physical work every year.

Google's new frontier model writes up to a million output tokens, where earlier Gemini models stopped at 64,000. Google's own agents used it to rewrite the libgav1 video decoder in safe Rust: 2.7 times faster, identical output.

Artificial Analysis clocked roughly 193,000 output tokens per task at max effort. That's the heaviest token use they have ever measured, and about seven times what GPT-6 Astra needs.

You'll summon a skill with a forward slash inside a task thread. An interface engineers tend to like and everyone else tends not to.

950 agents spent 21 hours combing through 200,000 reverse transcriptases. One of them wrote, mid-analysis: 'This is spectacular.'

One early tester pushed through a 680,000-line migration in under a day. Cache reads, the biggest line item in any agent bill, drop from 50 to 20 cents per million tokens.

On AutomationBench, Sol at xhigh effort beats Claude Opus 5 at max effort for nine percent of the cost per task. Luna comes in at 10 cents per million input tokens.

TypeSafe calls its yes/no question a Noul. The CEO confirmed on Hacker News that it is short for Bernoulli.

Xiaomi puts the entire reinforcement-learning run at $2.62 million for Pro and $850,000 for Flash, finished in under six days. More than half that money never touched the weights.

One line in the Fable 5.1 system card has gone almost unquoted these past two weeks: Anthropic sees no clear sign of dramatic acceleration.

Claude is meant to direct robots running experiments with little human intervention. Clinical trials are off the table for now.

Roughly 30,000 agents work on research and engineering at Anthropic at any one time. Every single action they take passes through a monitor first.

Designing a protein binder used to cost up to $10,000 per target. It now takes roughly $150 for GPUs and tokens combined.

The previous version was downloaded more than eleven million times. PrismML has still raised only $22.25 million.

39 cents per thousand workflows, against $3.31 for GPT-5.6 Luna — TypeSafe's own figure. Jev answers in under 100 milliseconds with a typed value instead of text.

97 languages it switches between mid-sentence, without being told. And tool calls that keep running in the background while the conversation carries on.

Terence Tao, Peter Scholze, Maryna Viazovska and 22 more holders of their field's highest honour have signed a joint declaration. The charge: using mathematics as a benchmark destroys the very thing mathematics is for.

100 teams, 40 hours, $20,000 in tokens each, and one question: how do you use AI responsibly to deepen human understanding of mathematics? After an open letter from Caltech, OpenAI is out as a sponsor. Half the money went with it.