Anthropic is considering rolling out a new AI model to counter OpenAI’s momentum since the launch of GPT-6 Astra, according to three sources cited by Reuters, a deliberation that sits awkwardly next to the company’s own safety messaging.
The potential launch comes ahead of Anthropic’s expected IPO and follows a September 12 essay in which CEO Dario Amodei called on the global AI community to slow the pace of releasing new capabilities to address safety concerns. “We must slow the pace at which we improve the capabilities of AI models,” Amodei wrote in the 3,800-word piece, which painted a picture of swarms of AI agents outpacing human control. OpenAI CEO Sam Altman and Elon Musk voiced support for the idea.
Two weeks later, the company is weighing exactly the kind of release its CEO asked the industry to pause. One person familiar with the matter told Reuters that Anthropic is evaluating the safety of its next model as part of the deliberation. No launch date has been announced, and the company has not confirmed the plan.
Why the competitive pressure is real
The commercial backdrop explains the tension. OpenAI released GPT-6 Astra on September 3, and the model has gained traction among business customers fast. According to spending data from corporate card platform Ramp, GPT-6 Astra now accounts for roughly 13 percent of enterprise AI spending tracked on the platform, compared with 8 percent for Anthropic’s Claude. Some investors have started re-evaluating Anthropic’s position as the leading provider of enterprise AI tools.
The revenue numbers are still on Anthropic’s side. Its annualized revenue run rate hit $65 billion by the end of July, up from $9 billion at the end of 2025, according to figures cited by Reuters. OpenAI’s run rate reached about $40 billion over the same period. But enterprise spending share is the metric that shapes the narrative ahead of a listing, and that is the one moving in OpenAI’s direction.
The IPO itself is shaping up to be one of the largest in history. Anthropic is in talks with Nvidia about an anchor investment of up to $10 billion in a listing that could raise as much as $100 billion at a valuation around $2 trillion, Reuters has reported. The company has reportedly considered timing the listing after the U.S. midterm elections in November, and its prospectus is expected to address investor concerns about public backlash to AI directly, according to CNBC reporting from August.
The slowdown debate, in practice
Amodei’s essay acknowledged the hardest version of his own argument: what happens if rival nations, particularly China, decline to slow down. He said the long-term goal would be coordinated international speed limits on frontier development, while admitting the incentives to pull ahead are enormous given potential military advantages. The proposal drew support from some peers and skepticism from others, with Cohere’s chief executive among those calling a proposed industry standards body a cartel.
That standards body is moving forward anyway. OpenAI said this week it is working with Anthropic and Google DeepMind on a self-regulatory organization modeled after FINRA, the body that oversees U.S. brokers and investment firms, following a July proposal by Google DeepMind’s Demis Hassabis to test the most powerful AI systems before public release. OpenAI policy chief Chris Lehane said the engagement had been under way for several weeks and that no antitrust waiver is needed for the firms to coordinate on safety matters.
Against that backdrop, Anthropic’s deliberation reads as a test of whether safety commitments survive contact with a competitive market. The company released Claude Fable 5.1 and Mythos 5.1 on September 1, so a new model would come weeks after its last frontier release, not months. Critics of the slowdown push have noted that frontier labs have endorsed coordinated pauses while continuing to train and deploy their own models, and this episode fits that pattern. The Future of Life Institute’s summer index gave Anthropic a C+, the best score among nine companies evaluated, which shows the company leads on safety governance even as it competes on speed.
There is also a workforce dimension. Anthropic said this week that Claude now leads 26 percent of its own AI research and development work, up from under 1 percent in February, a figure that doubles as a productivity claim and a data point in the debate over how fast capabilities are compounding. If models are increasingly doing the work of building their successors, the argument for deliberate pacing gets harder to sustain in both directions: the upside grows, and so does the risk of losing control of the process.
For enterprise buyers, the practical question is pricing and availability. GPT-6 Astra raised the frontier API rate to 2.5 times GPT-5.6 Sol’s level, and Anthropic’s September releases kept Fable pricing at $10 and $50 per million tokens while cutting cache-read costs by 75 percent, a meaningful discount on long agentic tasks. If Anthropic does launch a new model this month, expect the pricing to be part of the pitch, not an afterthought, because Ramp’s data suggests buyers are already migrating toward OpenAI’s newest flagship.
Neither company has announced anything. What is confirmed is the pattern: the same week the industry agreed to talk about slowing down, two of its three biggest players were preparing their next acceleration. The safety evaluation underway at Anthropic, per Reuters’ sources, will determine whether the gap between the message and the calendar closes or widens. For a company weeks away from what could be the largest IPO in history, that is not a philosophical question anymore. It is a valuation one.