AI
METR finds AI is doubling the pace of Anthropic's own model progress
An 11-week assessment published on 5 November by the safety nonprofit METR concludes that Anthropic is probably improving model capabilities on well-specified tasks at about twice the rate it managed between March 2024 and October 2025, with benchmark scores growing 1.9 times faster since October 2025, a 95 percent interval of 1.5 to 2.8 times. METR said it was confident the acceleration was not driven by more training compute and that most of it came from AI-assisted research, though it cautioned that Anthropic redacted much of the underlying evidence.