METR: AI roughly doubles Anthropic's pace of capability progress
According to METR, Anthropic is improving its models on well-specified tasks about twice as fast as between March 2024 and October 2025.
By GennnAi··1 min read
METR ran a high-access evaluation at Anthropic from August 12 to October 29, 2026. Benchmark scores (AECI, Anthropic's version of the Epoch Capabilities Index) are rising at 1.9 times the pre-October 2025 rate (95% confidence interval: 1.5–2.8). Most of the acceleration likely stems from AI use in research, not faster compute scaling. Anthropic redacted most of the key evidence, and METR acknowledges significant uncertainty. Further acceleration remains a significant risk.
Source: METR, published on November 5, 2026.