METR, an AI safety nonprofit also known as Model Evaluation and Threat Research, said on Thursday, November 5, that Anthropic is likely improving the capabilities of its models on well-specified tasks at around twice the rate it did between March 2024 and October 2025, according to an independent assessment based on high-access review inside the company from August 12 to October 29.
On Anthropic’s AECI benchmark index, METR estimated that scores have been growing at 1.9 times the pre-October 2025 rate since then, with a 95% confidence interval of 1.5x to 2.8x. That implies Anthropic made roughly two years of historical benchmark progress over the last year, METR said. The group stressed high uncertainty and said it would not be surprised if the true acceleration were closer to 1.3x or 3x.
METR said a majority of the current acceleration is likely attributable to Anthropic’s extensive use of AI to do R&D, estimating that AI may account for perhaps about 80% of the current acceleration while warning the figure is uncertain. Reuters reported those findings and said the study offers evidence that AI is accelerating its own development. METR said it was confident the acceleration was not due to faster scaling of training compute.
The assessment found no convincing evidence of major bottlenecks in algorithmic progress, training compute, experiment compute or data that would block further progress over the next 12 months, though METR said its view of bottlenecks changed several times and it is not confident in that assessment. METR also said AI is not yet making highly significant, fully autonomous contributions to the most important stages of model training, but that researchers described “signs of life” that such contributions could grow soon.
Anthropic redacted a majority of the evidence METR found most informative, describing the cuts as aimed at removing confidential information that could accelerate other AI developers, METR said. As a result, the public has to rely more on METR’s judgment than would be ideal, the group wrote.
AP reported that the analysis comes amid Anthropic CEO Dario Amodei’s calls to slow the technology’s pace of improvement. METR said Anthropic offered unprecedented access to internal R&D metrics, interviews and unreleased-model performance data, and that it took no payment from Anthropic.
Anthropic has previously published its own R&D Automation Index and said Claude led 26% of its AI R&D work. METR said its engagement was distinct from a separate independent-evaluator update Anthropic announced on October 15.
Sources
- METR: Independent Assessment of AI R&D Acceleration at Anthropic in October 2026 (November 5, 2026)
- Reuters via Investing.com: Anthropic models improving nearly twice as fast with help from AI, study finds (November 5, 2026; Reuters byline confirmed)
- AP: Anthropic is rapidly accelerating AI development amid slowdown calls, outside research group says (November 5, 2026; fetch limited, verified via AP headline/lede and secondary carries)
- The Independent: Anthropic AI models are improving twice as fast as this time last year, by Anthony Cuthbertson (November 6, 2026)
- AI Tech Daily: What Anthropic’s R&D Automation Index measures (September 24, 2026)
- AI Tech Daily: Anthropic says Claude leads 26% of its AI R&D work (September 18, 2026)