GPT-6 Astra ships as Nvidia confirms Hugging Face
Astra started rolling out on Thursday at $10 and $50 per million input and output tokens, two and a half times GPT-5.6 Sol's current rate, and Jensen Huang confirmed Nvidia is buying Hugging Face for $12.93bn a week after the deal was first reported as agreed.
OpenAI shipped GPT-6 Astra on Thursday OpenAI began rolling Astra out to enterprise customers in its gated Daybreak programme, with ChatGPT Plus, Pro, Business and Enterprise, the API and AWS to follow over the coming days. It costs $10 per million input tokens and $50 per million output — two and a half times GPT-5.6 Sol’s current promotional rate, with a Fast mode at $20 and $100 — and it is the same model OpenAI rated a Critical cyber risk under its own Preparedness Framework a day earlier. Source
Nvidia confirmed it is buying Hugging Face for $12.93bn Jensen Huang confirmed the acquisition on Thursday, a week after The Information reported it had been agreed and Bloomberg reported that nothing had been signed. Hugging Face hosts around three million models and half a million datasets for some 18 million developers, and Huang says it “will remain an open platform for the entire AI ecosystem”; it is Nvidia’s second-largest purchase after the $20bn it paid for Groq’s assets last year. Source
Anthropic will make its IPO prospectus public after Labor Day The Information reported on Thursday that Anthropic plans to publish its S-1 once the US holiday passes on 7 September, hold an investor day in the middle of the month, and list in late September or early October. The June filing was confidential and the company has released no financials, so the $10.9bn second-quarter revenue and $559m operating profit being quoted around the story are analysts’ estimates rather than disclosure. Source
The Pentagon says its Anthropic ban is still in force Under Secretary of Defense for Research and Engineering Emil Michael said on Thursday that the “supply chain risk” designation remains in effect, a day after Commerce Secretary Howard Lutnick told Axios that Anthropic is “back on the right side” and that “we trust Anthropic”. A federal judge vacated the designation on 27 August as unlawful retaliation for criticising the government, and ChatGPT Mil and Grok for Government went live on the Pentagon’s GenAI.mil platform on 31 August. Source
ChatGPT, Claude and Grok all went down at the same time Downdetector reports spiked within minutes of each other on Thursday morning US time, taking down dependent tools including Cursor. OpenAI attributed its outage to a routing error from about 7:43am Pacific that made ChatGPT and Codex unavailable for some users, Anthropic said its Opus models were the last to recover, and all three services were back to normal by early afternoon Pacific. Several outlets noted a concurrent Microsoft Azure incident as a possible common factor; none of the three companies has confirmed a shared cause. Source
Moonshot AI filed confidentially for a $3bn Hong Kong IPO The Kimi K3 maker filed on Thursday, valued at $50bn in its current funding round, having first redomiciled from its offshore holding structure into mainland China, which is a prerequisite for the listing. That is two frontier labs moving towards a public market on the same day, on opposite sides of the export-control line. Source
Meta shipped Muse Spark 1.3 on Thursday, still proprietary The fourth Muse Spark release in five months went to Muse Code and the Meta Model API at $1.25 and $4.25 per million input and output tokens, scoring 75.4 on DeepSWE 1.1 against 1.2’s 59.3 on Meta’s own launch table, with a contributor tier at $0.10 and $0.20 for developers who let Meta train on their prompts and completions. Every number in Meta’s launch table comes from a “max” variant that is in limited preview and still in safety testing, and Zuckerberg’s post the same day promising the open-weights release “soon” no longer names a version, a date, a model size or a licence. Source