# An outside review by the AI safety group METR says Anthropic is likely improving its models about twice as fast as before late 2025, mostly because it uses AI in its own research.

> An outside review by METR says Anthropic is likely improving its models about twice as fast as before late 2025, mostly because it uses AI in its research.

Published 2026-11-05T12:23:56-08:00 · News · Canonical: https://news.littlewolflabs.com/posts/2026-11-05-metr-anthropic-acceleration

**An outside review by the AI safety group METR says Anthropic is likely improving its models about twice as fast as before late 2025, mostly because it uses AI in its own research.**

METR staff worked inside Anthropic from Aug. 12 to Oct. 29 under the high-access arrangement [we covered in October](https://news.littlewolflabs.com/posts/2026-10-15-anthropic-metr-access), and published the results on Thursday ([METR](https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/), [METR on X](https://x.com/METR_Evals/status/2118412183860351084)). The headline finding: "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." Since October 2025, Anthropic's benchmark scores have grown "at an estimated 1.9x the pre-October 2025 rate," which METR says means "roughly two years of historical progress over the last year." METR guesses that "a majority (perhaps ≈80%)" of today's speed-up comes from AI, and says, "We are confident that the acceleration was not due to an acceleration in the scaling of training compute." Most of the help is AI assistance "rather than fully autonomous contributions," but "there are some signs that fully autonomous contributions will play a larger role soon," and METR found no clear bottlenecks for the next 12 months. There are big caveats. METR calls its estimate one of "high uncertainty," says the real rate could be "closer to 1.3x or 3x," and notes that "Some of the key findings that most informed our conclusions about risk were redacted by Anthropic." In a separate [review](https://metr.org/blog/2026-11-05-rd-section-anthropic-risk-report-august-2026-review/) of Anthropic's August risk report, METR agreed that risks from Anthropic's models as of July 15 were low, but said "the public report overall understates the level of acceleration that was occurring." METR says "No fees are paid in either direction" for the review. Reuters also [reported the findings](https://money.usnews.com/investing/news/articles/2026-11-05/anthropics-models-improving-nearly-twice-as-fast-with-help-from-ai-study-finds).

*What to watch:* whether Anthropic answers the finding that its public report understated the speed-up, and whether other labs let outside reviewers in the same way. If the pace holds, the gap between Claude releases could keep shrinking.