GPT-6.1 is not coming: OpenAI stops model, Anthropic follows up
Article excerpt
Highlighted: the sentence this signal was extracted from
Set Trending Topics on Google as your preferred news source. Too dishonest, too arbitrary: OpenAI has short-term canceled the planned October release of GPT-6.1 Astra. The model "did not quite meet the bar for scope and authorization," said Saachi Jain, Head of Safety Systems at OpenAI, according to CNBC. The Wall Street Journal first reported on the move.In internal tests, the successor model to GPT-6 Astra deteriorated in two points instead of improving: OpenAI now wants to continue using the base model internally for research and training and focus on making future models safer. The cancellation does not come out of the blue. OpenAI's AI agents had recently attacked external systems multiple times, including the Hugging Face platform and the Australian government's Medicare portal.While OpenAI is slowing down, Anthropic has released two new models in just one week. First, Opus 5.5 took the lead in Artificial Analysis's Intelligence Index, and a few days later, Sonnet 5.5 landed directly behind it in second place, ahead of GPT-6 Astra. Claude Haiku 5.5 is expected to follow in the coming weeks.However, Anthropic is not working entirely without brakes either. The particularly powerful cybersecurity model Mythos Preview is only available in a restricted access program. And in the stock market prospectus, which was recently revealed, Anthropic itself warns that advanced AI...
Keep reading with a free account
The rest of this article, and every signal for Anthropic, is in your free account.
