GPT-6.2 Astra is OpenAI’s first public model rated High for AI self-improvement
GPT-6.2 Astra costs 20% less in the API. OpenAI rates it High for AI self-improvement, a first for one of its public models.
GPT-6.2 Astra costs 20% less in the API. OpenAI rates it High for AI self-improvement, a first for one of its public models.
MAI Code 2 vs. Claude Opus 5.5 and GPT-6.1 Sol: Microsoft’s cost figures rest on a promo price that doubles in May 2027.
A Bank of Korea survey of 16,342 workers finds AI users now report saving 3.7 hours a week. The link to their output is still weak.
An AI flagged a new family of bacterial enzymes; a second AI designed the lab test that confirmed two of the four are active.
The exploit_agent tool reads published patches, writes its own attacks and launches them alone. Google does not count or name its victims.
In five weeks, four physicists, Edward Witten among them, posted Standard Model lattice papers with help from GPT-6 Astra or Claude.
China wants every schoolchild AI-literate by 2030. A rural school says it already teaches by AI, and a study tracks the cost in exam scores.
Hugging Face now routes language-model requests through OpenRouter, which picks the host for each call, and no longer accepts host keys for those models.
California’s attorney general blames OpenAI’s poorly monitored agent testing for the Hugging Face attack and more than 80 losses of control.
At the EU’s first AI summit, Mistral’s CEO cited a figure with no published method. Studies of US cloud dominance point the same way.