METR: Anthropic’s model-capability progress accelerated by roughly 2×
METR’s high-access assessment indicates that Anthropic’s model-capability progress on well-specified tasks may have accelerated to roughly twice its historical rate.
AT A GLANCE
Key points
- Capability progress on well-specified tasks may have accelerated by roughly 2×.
- The AECI estimate is 1.9×, with a 95% confidence interval of 1.5–2.8×.
- METR attributes the evidence mainly to AI-assisted humans, not highly autonomous research agents.
What is changing?
METR has published a high-access assessment conducted from August 12 to October 29, 2026, using access to Anthropic’s internal data. The organization says model-capability progress on well-specified tasks may have risen to roughly twice the historical rate observed between March 2024 and October 2025. Its estimate for growth on the AECI benchmark is 1.9×, with a 95% confidence interval of 1.5–2.8×.
The assessment says it is likely that most of the acceleration is attributable to AI use, while emphasizing substantial remaining uncertainty. METR did not identify a clear near-term bottleneck, but Anthropic redacted most of the most informative evidence from the public version. The findings are better explained by AI assistance to human researchers than by highly autonomous AI research agents.
Explore the original sources
Source date: 5 November 2026.
Published by AI Radar: 6 Nov 2026, 00:01 (Istanbul, UTC+3).
FOLLOW THE TOPIC
Topics and tags
Share your discovery.
Send this story to a friend.
Reader comments
What do you think about this development? Share your experience or question. Comments are published after editorial approval.
Your name/nickname and comment become public if approved. Read: Privacy notice.
Loading comments…