From March 2024 through October 2025, Anthropic’s progress offered a benchmark for how quickly a frontier AI lab could improve its models. The new assessment argues that the pace has since changed dramatically.

METR, an AI-safety organization that examined Anthropic’s research trajectory, concluded the company is “likely improving the capabilities of their models on well-specified tasks at around twice the rate they did between March 2024 and October 2025.” It said the acceleration was likely tied to Anthropic’s extensive use of AI in its own research and development.¹

The timing makes the conclusion especially awkward for Dario Amodei. METR began discussing the assessment with Anthropic in mid-August, according to the report’s account. Then, in mid-September, Amodei published an essay titled “We Must Pace the Frontier” — a public case for slowing and managing the advance of the most capable systems rather than treating speed as an unquestioned virtue.¹

METR submitted its final public report on November 3. When the analysis was released Thursday, the headline finding was straightforward: Anthropic was developing AI systems at roughly twice the speed of a year earlier, even as its CEO warned against an uncontrolled race.²

That does not necessarily mean Amodei’s call for restraint is insincere; the assessment describes observed capability gains, not the company’s preferred policy. But it underscores the central tension in frontier AI: a lab can advocate caution while deploying tools that make its own discovery engine faster. For Anthropic, AI-assisted R&D may be proving both the promise of the technology and the difficulty of pacing it.