Reuters reported on September 19, citing three sources, that Anthropic is considering releasing a new model to counter competitive pressure after OpenAI launched GPT-6 Astra, with the timing falling ahead of the company's expected IPO. The same sources said the model's safety is also being evaluated internally. Anthropic has not commented publicly.
The only figure in the report with any real backing comes from enterprise spend-tracking platform Ramp's latest data: Astra accounts for about 13% of the enterprise AI spending it observes, versus roughly 8% for Anthropic's Claude Fable. That is a sampled measure covering reimbursements and subscription transactions from Ramp's own customer base — it is not equivalent to either company's revenue share, and it does not capture self-hosted deployments.
Colliding With a Statement From a Week Earlier
On September 12, Anthropic CEO Dario Amodei published a roughly 3,800-word essay calling on the entire industry to slow the pace of capability gains and leave more room for safety alignment.
We must slow the pace at which we improve the capabilities of AI models.
Seven days later comes word that the same company is considering rushing out a targeted new model. The two moves aren't necessarily in conflict at the corporate level: what's meant to slow down is the push on the upper bound of capability, while what's being rushed out may just be an engineering-focused version of existing capability. What's unclear is that outsiders have no way to judge — Anthropic hasn't said which tier the new model falls into, or even confirmed it exists.
The IPO Timeline Keeps Slipping
Two people familiar with the matter said Anthropic may push its IPO to after the U.S. midterm elections in November, though the election itself isn't expected to significantly affect the offering. Terms that circulated earlier pointed to a valuation of around $2 trillion and a raise of up to $100 billion; in mid-September there were also reports that Nvidia was considering subscribing for as much as $10 billion in the round. None of these figures has been confirmed by the company, and the underwriting syndicate, listing venue, and share count remain undisclosed.
Lined up across three weeks, the sequence goes: first an open letter calling for the whole industry to slow down, then a strategic subscription widely read as securing a front-row seat, and now a model meant to fend off a rival. Each step holds up on its own, but read together, the roadshow will likely run into the same question: once the slow-down pledge is written into the prospectus, who decides the pace of model releases?
Competitive Pressure Is Real; the Share Numbers Are Shaky
Ramp's 13% versus 8% keeps getting cited because it's one of the few third-party spend windows that updates monthly. Its weaknesses are just as clear: it only looks at credit-card and reimbursement transactions, leaving out direct API connections, private cloud deployments, and channel resale. Neither company has published comparable enterprise-side revenue figures, so any share estimate built on these numbers should be treated with caution.
The next verifiable milestone is specific: whether Anthropic releases a new model around the time it files for the IPO, and whether that model's capability tier lands above or alongside Claude Fable 5.1. Until then, every line in the report carries only sourcing from unnamed people familiar with the matter.
Sources: Reuters, Ramp enterprise spend tracker, CocoLoop, Seeking Alpha. Valuation and raise figures are cited from previously reported terms and have not been confirmed by the company.