METR, the AI evaluation nonprofit, published an independent assessment on November 5, 2026 of how much AI is accelerating research and development at Anthropic. METR staff carried out the evaluation between August 12 and October 29, 2026, with access to Anthropic's internal data.
METR's best guess is that, as of October 2026, Anthropic is improving its models on well-specified tasks at roughly twice the rate it did between March 2024 and October 2025. The report attaches high uncertainty to that figure, saying METR "would not be surprised if the current rate of growth was closer to 1.3x or 3x the historical rate." A separate measure based on Anthropic's internal benchmark index, AECI, shows 1.9x acceleration since October 2025, with a 95% confidence interval of 1.5x to 2.8x.
METR estimates that "a majority (perhaps ≈80%)" of the current acceleration is due to AI, and says it is confident the speed-up did not come from faster scaling of training compute. Staff concluded that R&D is accelerated "primarily via AI assistance rather than fully autonomous contributions," but reported "signs of life" that fully autonomous contributions could grow soon. METR found no convincing bottlenecks to continued rapid progress over the next 12 months, while noting that it is "not confident" in that assessment.
Anthropic gave METR internal benchmark scores, historical algorithmic efficiency and training compute data, headcount and lines-of-code data, internal uplift surveys, agent inference spend, and about 15 staff interviews. METR did not receive query access to unreleased models. Anthropic held unilateral redaction rights and redacted a majority of the evidence METR found most informative, citing confidential information that could accelerate other developers. Anthropic also cancelled a planned compute-allocation verification component.
This appears to be the first third-party, high-access attempt to quantify AI-driven R&D acceleration inside a frontier lab, but the redactions limit how much outsiders can check. METR also published a companion review of the automated R&D section of Anthropic's August 2026 Risk Report. That review agrees that direct catastrophic risk from automated R&D was low as of July, but argues that the public report signals more confidence about AI's share of the speed-up than the private evidence supports.