Anthropic is considering shipping a new model to answer OpenAI, Reuters reports, citing three sources. The trigger is GPT-6 Astra, which OpenAI released on September 3 and which is landing well with businesses.
Astra is catching up, and you can measure it
Two numbers sit in the report. On the expense platform Ramp, Astra now accounts for roughly 13 percent of tracked corporate AI spending, against about 8 percent for Claude Fable. And on OpenRouter, where developers spread their traffic across models, users spent more with OpenAI than with Anthropic last week — the first time in more than two and a half years.
One source says Anthropic is evaluating the safety of its next model, and that is part of the release decision. Alongside it runs an internal argument about how much money goes into new models and how much into getting profitable. Rising rates have made investors impatient. Anthropic declined to comment.
On revenue, Anthropic is still ahead
Annualized revenue passed $65 billion at the end of July, up from around $9 billion at the end of 2025. The company projects $190 billion to $200 billion for 2028. OpenAI was past $40 billion in July. Some investors therefore see Astra as no real threat: once you are inside a large enterprise, you do not get pushed out quickly.
They see the bigger danger elsewhere. Open-weight models push token prices down and invite companies to build more themselves instead of buying. Meta is one of Anthropic’s largest customers and wants to cut its own usage, according to Reuters.
The contradiction is in the essay itself
On September 12, Dario Amodei wrote across 3,800 words: ‘We must slow the pace at which we improve the capabilities of AI models.’ Six days later the company is weighing whether to pull a model forward because a rival is taking share. Both can be true — you can believe in slowing down and still have to ship. But a call for restraint reads differently when it comes from the house that is currently in front.
The IPO does not make it easier. It has moved to November, possibly past the midterms. Until then every investor is watching whether Anthropic holds its lead with enterprise customers. A new model would be the fastest answer — and the worst illustration for that essay.