According to Reuters, three sources said that Anthropic is considering launching a new artificial intelligence model to counter the strong momentum of OpenAI's release of GPT-6 Astra. This comes as the company prepares for its IPO - just before this, its CEO had also called on the entire AI industry to slow down the pace of technological iteration.

The sources said that launching a new model is intended to face pressure from direct competitors. Previously, Anthropic CEO Dario Amodei had spoken out to the global AI industry, urging the industry to slow down the release of new capabilities and properly address safety risks. "We must slow down the speed of AI model capability iteration," he wrote in an article of 3,800 words on September 12. The article depicted a bleak outlook: a large number of AI agents are sweeping the internet, with development speeds exceeding human control, a view supported by Altman and Musk.

Astra Surpasses Claude in the Enterprise Market

OpenAI's GPT-6 Astra, released on September 3, emphasized computer operations, software engineering, cybersecurity, and professional office capabilities, and has quickly gained traction among business customers. Data from the enterprise expense management platform Ramp shows that Astra now accounts for about 13% of enterprise AI-related spending, while Anthropic's Claude Fable is around 8%. On OpenRouter, last week users spent more on OpenAI models than on Anthropic models, marking the first time OpenAI has led in this metric in two and a half years. As a result, some investors have reassessed Anthropic's position as a leading enterprise AI supplier.

However, Anthropic still maintains a revenue advantage: as of the end of July, its annualized revenue run rate has exceeded $65 billion (approximately RMB 436.696 billion), while it was only about $9 billion by the end of 2025; the company predicts its revenue will reach between $190 billion and $200 billion by 2028. OpenAI's annualized revenue run rate in July exceeded $40 billion.

Bigger Challenges: Rise of Open Source Models and Meta's Shift to In-House Development

Investors believe that a bigger potential threat comes from the rise of open-source, open-weight models - these models can reduce token call costs and allow companies to build their own AI infrastructure. A source said that Meta, which was originally one of Anthropic's most important clients, is planning to reduce its use of its models and instead develop more AI capabilities internally.

A source revealed that while Anthropic is discussing whether to release a new version, it is still assessing the safety level of this next-generation model; the company is also discussing how to balance new model development with improving profitability. Regarding the report, Anthropic declined to comment.

Regarding the IPO schedule, Altman has confirmed that OpenAI will not launch an IPO in 2026; according to two sources, Anthropic may delay its IPO until after the US November mid-term elections, previously Reuters reported that it was expected to start the IPO roadshow as early as mid-October.