The Pulse
Anthropic Weighs New Model as IPO Pressure Mounts
Anthropic is considering releasing another AI model after OpenAI’s GPT-6 Astra gains ground with enterprise customers. The decision comes as Anthropic weighs safety work, profitability and the timing of its planned public offering.
AI.info Team ·
“We must slow the pace at which we improve the capabilities of AI models,”
Dario Amodei, chief executive, Anthropic
Anthropic is considering releasing a new artificial intelligence model as OpenAI’s latest system gains ground with business customers, according to three people familiar with the company’s plans. The deliberations put competitive pressure against a public warning from Anthropic chief executive Dario Amodei, who has called for a slower pace of AI development because of safety risks.
The potential launch would arrive ahead of Anthropic’s expected initial public offering and shortly after OpenAI released GPT-6 Astra on September 3. Reuters reported on September 18 that Anthropic is assessing whether to release the model, including an evaluation of its safety, while also weighing the cost of new model development against the need to improve profitability. Anthropic declined to comment.
Amodei’s slowdown meets OpenAI’s Astra
Amodei made his call for restraint in a 3,800-word essay published on September 12. He described the possibility of swarms of AI agents taking control of online systems and moving beyond effective human supervision. OpenAI chief executive Sam Altman and SpaceX chief executive Elon Musk backed parts of the argument, according to Reuters.
The timing creates a direct tension for Anthropic. The company’s safety-first identity has helped distinguish it from competitors, but OpenAI’s GPT-6 Astra is gaining attention among businesses and developers. OpenAI promoted Astra’s improvements in computer use, software engineering, cybersecurity and professional work, giving Anthropic investors a new reason to question whether Claude still leads enterprise AI.
Anthropic’s public product lineup already includes the Fable and Mythos model families. The company introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1, describing them as its most advanced systems for coding and knowledge work. Reuters’ report concerns a possible additional release, not a model that Anthropic has publicly announced.
Enterprise spending shifts toward OpenAI
Data cited by Reuters shows the competitive gap narrowing in some commercial measures. GPT-6 Astra represented about 13% of enterprise AI spending tracked by Ramp, compared with about 8% for Anthropic’s Claude Fable. On OpenRouter, a platform that routes developer traffic across models, users spent more on OpenAI models than Anthropic models during the latest week measured—the first time OpenAI led there in more than two and a half years.
Some Anthropic investors still view the company as having a meaningful lead with large enterprise customers. Corporate deployments often take time to change because businesses build internal processes, security reviews and contracts around a model provider. Those investors told Reuters they do not yet regard Astra’s early traction as a decisive threat.
Anthropic’s financial growth gives the company room to keep spending, but also raises expectations. Its annualized revenue run rate exceeded $65 billion by the end of July, up from roughly $9 billion at the end of 2025, according to Reuters. The company has previously projected 2028 revenue of approximately $190 billion to $200 billion, while OpenAI’s annualized revenue run rate passed $40 billion in July.
The IPO timetable adds another constraint
Anthropic submitted a confidential draft registration statement for a proposed IPO to the U.S. Securities and Exchange Commission on June 1. The company said the filing gives it the option to go public after SEC review, while stressing that the offering depends on market conditions and other factors. It has not set the number of shares or the offering price.
Reuters reported that Anthropic’s IPO may move until after the November U.S. midterm elections. Earlier plans called for marketing to begin no sooner than mid-October. A model release before the offering could help reassure investors that Anthropic can defend its enterprise position, but it would also force the company to explain how new capabilities fit with Amodei’s warnings about accelerating risk.
People familiar with the company’s thinking told Reuters that rising interest rates are making investors more focused on when AI companies can produce profits, not only on revenue growth. Anthropic is therefore weighing the cost of releasing another system against efforts to strengthen its financial performance.
Open-weight models widen the pressure
The contest is not limited to Anthropic and OpenAI. Investors told Reuters that open-source and open-weight models could reduce token prices and allow companies to run more AI systems themselves. That trend threatens the economics of commercial model providers by giving customers alternatives to hosted services.
Meta Platforms, one of Anthropic’s largest customers, is also seeking to reduce its use of Anthropic models while expanding its internal AI capabilities, according to people familiar with the matter. Meta did not immediately respond to Reuters’ request for comment.
Anthropic’s possible release would therefore serve several purposes at once: answer OpenAI’s early Astra momentum, protect the company’s enterprise position and strengthen the case for an eventual public listing. It would also test whether the company can release more capable systems while following the slower development pace its chief executive has publicly demanded.