Anthropic is considering launching a new artificial intelligence model. It wants to respond to rival OpenAI, which introduced a new model called GPT-6 Astra in early September and is rapidly gaining customers with it. According to Reuters and three of its sources, the move is being discussed internally shortly before Anthropic plans to offer its shares on the stock market. Management is hesitating over whether to release another model now, as the competition is hot on its heels. One source added that the company is still examining whether the new model is sufficiently safe and will make a decision based on the results.
Astra is taking Anthropic's paying users
When introducing Astra, OpenAI emphasized that the model is better at operating computers, programs better and faster, has greater cybersecurity capabilities, and handles professional work more effectively. Companies and developers have responded well to the new model, which has begun to worry investors looking to put money into Anthropic. They are asking whether OpenAI can win over customers among large enterprises, where Anthropic had held the upper hand for many months.
The data so far suggests that it can. Ramp tracks corporate spending and found that roughly 13 percent of the money spent on artificial intelligence went to Astra, compared with approximately eight percent for Anthropic's Claude Fable model. OpenRouter, a service through which developers select specific models, reported similar results. Last week, users there spent more money on OpenAI models than on Anthropic's, which happened for the first time in more than two and a half years.
Spend on development or start making money
The company is also discussing internally how much money to invest in developing new models and how much to set aside so that Anthropic can become profitable sooner. Banks have made borrowing more expensive, and investors are now more interested in when they will see their first returns.
Pressure is also coming from other directions. There are models that anyone can download and run themselves, often at a significantly lower cost. China holds a strong position in this area. Large companies can therefore build their own solutions instead of paying Anthropic or OpenAI. According to investors, this could significantly affect prices across the industry, as customers suddenly have far more options.
One such case is already emerging. Meta is one of Anthropic's largest customers, but according to people familiar with the discussions, it wants to use Anthropic's models less and rely on its own development. Meta declined to comment on the matter.
Stock market debut may wait until after the U.S. elections
Not everyone, however, sees Astra as a major threat. Some of Anthropic's current and prospective investors point out that its lead among large companies remains substantial. After all, switching from a provider that an entire company is accustomed to takes a long time.
Revenue figures support this as well. By the end of July, Anthropic was reporting annualized revenue of more than $65 million, compared with around $9 million at the end of 2025. For 2028, it expects approximately $190 million to $200 million. OpenAI surpassed $40 million in July. Investors also expect the rankings among Anthropic, OpenAI, Google, and others to shift with each new generation of models, meaning no one will maintain a permanent lead.
Meanwhile, the rush to go public has eased somewhat. Altman confirmed on Saturday that OpenAI will not offer its shares publicly this year because concerns about artificial intelligence safety make it an inappropriate time to do so. Anthropic may postpone its stock market debut until after the November U.S. elections, which, according to two sources, should not have a major impact on it. The date had already been pushed back earlier, with the company expected to begin approaching investors no sooner than mid-October.
Sources: reuters.com and newsbreak.com
