📰 Anthropic has just publicly called for slowing down AI development, but just seven days later it was reported that the company is already discussing releasing a new model earlier. Honestly, that turn is pretty direct: its safety position hasn’t even finished being laid out, and enterprise customers have already “voted with usage.”

🔥 The pressure is coming from OpenAI’s Astra. About two weeks after it launched, Astra captured roughly 13% of enterprise AI spending on OpenRouter, while Anthropic’s Fable series was around 8%. In the first week of September, OpenAI’s weekly token usage share rose to about 65%, while Anthropic dropped to roughly 35%—and its leading position from the past two and a half years was overtaken.

Fable 5.1 also took action: when it was released on September 1, the price for cached reads was cut by 75%. But after Fable 5 launched two months earlier, its share hovered around 11% for a long time, and more customers chose Opus 5, which offers better value for money.

👀 Even more subtly, Anthropic’s IPO plans are reportedly possibly being pushed back to November. As the listing window gets closer, both revenue and customer share need to show results—so releasing a model early becomes a realistic option. The report also said that Claude is heavily involved in next-generation model development.

🤔 If an early release is true, do you think it’s a normal business response—or will it weaken the safety stance Anthropic has consistently emphasized?

#Anthropic #OpenAI #人工智能 #科技竞争