METR: Anthropic’s model progress likely ~2x historical pace, mostly from AI in R&D

METR posted on X on 5 November 2026 that it is publishing results from a high-access review inside Anthropic on whether AI itself is accelerating AI development:

“Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development. We are now publishing a report summarizing the results of this risk assessment.”

What METR’s report claims

In its 5 November 2026 assessment, METR says staff worked at Anthropic from 12 August to 29 October 2026 with access to internal R&D metrics, internal-model benchmarks, AI speedup analyses and staff interviews. METR’s subjective conclusions, as stated in the executive summary, include that as of October 2026 Anthropic 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,” with high uncertainty (METR says it would not be surprised by a true figure closer to 1.3x or 3x).

METR also says a majority of “this acceleration today is likely attributable to Anthropic’s extensive use of AI to do R&D,” while noting substantial uncertainty and that other causes likely mattered more in late 2025. Empirically, METR reports that since October 2025 Anthropic’s AECI benchmark scores have grown at an estimated 1.9x the pre-October 2025 rate (95% confidence interval 1.5–2.8x), and that the acceleration “was not due to an acceleration in the scaling of training compute.” METR says it did not see evidence of significant near-term bottlenecks in algorithmic progress, training compute, experiment compute or data over the next 12 months, while warning its bottleneck picture changed several times as more detail arrived.

Associated Press and Reuters coverage of the same report summarizes METR’s ~2x pace finding and the claim that AI assistance, not faster compute scaling, explains much of the speedup. Reuters relays METR’s rough guess that AI may account for perhaps 80% of the current acceleration, with the same uncertainty METR flags in its own write-up, and notes METR’s caution that AI is not yet making highly significant, fully autonomous contributions at the most important training stages.

Caveats

All quantitative claims above are METR’s (or secondary reporting of METR’s). Anthropic’s internal data were not independently re-measured for this article. METR itself stresses uncertainty bands and redactions. Treat the ~2x / ~80% figures as METR’s stated judgments, not confirmed ground truth.

This report is not legal, financial, or health advice. See the disclaimer.