Anthropic weighs a new model before its IPO as GPT-6 Astra takes enterprise share
Reuters, three sources: the lab is deliberating a release to counter OpenAI’s September 3 model. Ramp: Astra ~13% of tracked enterprise AI spend, Claude Fable ~8%. IPO could slip past the November midterms. Anthropic declined to comment.
Deep dive2 minIn-depth
Maya Chen2 min read
- Anthropic
- IPO
- GPT-6 Astra
- Claude Fable
- OpenRouter
- Safety

Photo: TechCrunch / Wikimedia CommonsCC BY 2.0
Anthropic is considering rolling out a new AI model to counter OpenAI’s GPT-6 Astra, three people familiar with the matter told Reuters, as the company prepares for an expected IPO and after its chief executive asked the industry to slow the pace of new capabilities. Echo Wang, Krystal Hu and Milana Vinn filed the exclusive late Friday New York time, published September 19. Astra launched on September 3. Anthropic declined to comment. The sources said the lab is evaluating the safety of the next model as part of the release deliberations. That is not a launch date. It is a live argument inside a company that spent the last week telling the rest of the industry to wait.
The pressure is in the enterprise numbers, not in a keynote. Ramp, the corporate expense platform, now shows Astra at about 13 percent of the enterprise AI spending it tracks, against about 8 percent for Anthropic’s Claude Fable, Reuters wrote. On OpenRouter, which routes developer traffic across models, users spent more on OpenAI than on Anthropic last week — the first time OpenAI has led that measure in more than two and a half years. Some potential IPO investors told the wire they are re-evaluating whether Anthropic is still the default enterprise vendor. Other existing and prospective investors in both listings said they do not think Astra is a serious threat yet, given how long it takes to unseat an incumbent inside a large company.
We must slow the pace at which we improve the capabilities of AI models.
— Dario Amodei, “We Must Pace the Frontier,” Sep 12, 2026, as quoted by Reuters, Sep 19, 2026
A $65 billion run-rate, and a calendar that keeps sliding
Anthropic still has the larger book. Its annualized revenue run-rate topped $65 billion by the end of July, Reuters wrote, up from about $9 billion at the end of 2025. The company is projecting 2028 revenue of roughly $190 billion to $200 billion, the wire said it had previously reported. OpenAI’s run-rate passed $40 billion in July. Leadership among Anthropic, OpenAI, Google and the rest will shift as each generation ships, investors told Reuters; any single-model lead is perishable. The more durable worry they named is open-weight competition, especially from China, which can cut token costs and let customers run more of the stack themselves.
Meta has been among Anthropic’s largest customers and is looking to reduce that use as it builds more capability in-house, people familiar with the matter said. Meta did not immediately respond. Some of the internal debate, Reuters wrote, is how to balance spending on a new model against strengthening profitability as rising rates make investors more interested in when cash shows up. Altman confirmed on Saturday that OpenAI would not go public in 2026, citing safety concerns, the exclusive said. Anthropic could push its own IPO to after the November U.S. midterms, two people familiar with the matter said. The elections are not expected to have a major effect on the offering. Marketing had already been expected to begin in mid-October at the earliest. A new model before a listing would test whether Anthropic can answer Astra without walking back the safety-first identity it has sold to those same investors.


