METR finds Anthropic is improving its models at about twice its earlier rate, guesses AI is doing most of it, and says Anthropic redacted much of the evidence
After about 11 weeks with Anthropic's internal data, the outside evaluator puts the speed-up at 1.9x and says Anthropic's August risk report understated it.
METR, the outside AI evaluation group, published two reports on 5 November on how fast Anthropic is improving its own models.
The main one, “Independent Assessment of AI R&D Acceleration at Anthropic in October 2026”, concludes that Anthropic is “likely improving the capabilities of their models on well-specified tasks at around twice the rate” it managed from March 2024 to October 2025. Its central estimate, built on Anthropic's own version of the Epoch Capabilities Index, is that benchmark scores have grown since October 2025 at 1.9 times the pre-October 2025 rate, with a 95% confidence interval of 1.5 to 2.8 times.
On the cause, the authors write: “We guess that a majority (perhaps ≈80%) of the current (October 2026) acceleration is due to AI.” METR worked from mid-August to late October, using internal benchmark scores, surveys, interviews and code metrics. It had no query access to unreleased models. Anthropic had the right to redact, and METR says: “Anthropic redacted a majority of the evidence we found most informative for these conclusions.”
A second METR report reviews the automated-research section of Anthropic's August 2026 risk report. It says “the public report overall understates the level of acceleration that was occurring”, and that it is plausible Anthropic's early-warning 2x threshold had already been crossed by 15 July.
The AP reported that Anthropic co-founder Sam McCandlish replied on X that many of METR's conclusions match his own.
- Confirmed METR's central estimate is that Anthropic's benchmark scores have grown since October 2025 at 1.9 times the pre-October 2025 rate, with a 95% confidence interval of 1.5 to 2.8 times. METR
- Confirmed METR guesses about 80% of the October 2026 acceleration is due to AI, and says Anthropic redacted a majority of the evidence it found most informative. METR
- Confirmed METR says Anthropic's August 2026 public risk report overall understates the acceleration that was occurring. METR
- Reported Sam McCandlish replied on X that many of METR's conclusions match his own, the AP reported. AP, via ClickOrlando
Safety, security & governanceScience & research
Today in the November 6, 2026 edition · front page