Anthropic's artificial intelligence models have advanced at approximately twice the rate observed in the preceding period, according to a report published Thursday by the Model Evaluation and Threat Research (METR) organization. The independent evaluation, conducted between August 12 and October 29, 2026, concluded that Anthropic's progress on well-specified tasks since October 2025 was equivalent to two years of development at its prior pace.METR, a nonprofit focused on AI safety, gained access to Anthropic's internal data, including benchmark scores for unreleased models, staff interviews, and various research and development metrics. The report indicates that the acceleration was not due to an increase in computing power. Instead, METR's subjective estimate suggests that approximately 80% of this acceleration stems from Anthropic's extensive use of AI in its own research and development processes.The findings provide some of the clearest evidence to date that AI is accelerating its own development, a process often referred to as recursive self-improvement. While the report notes that AI is not yet making fully autonomous contributions to the most critical stages of model training, it highlights "significant signs of life" that this could occur soon. Anthropic's co-founder, Sam McCandlish, acknowledged that many of METR's conclusions align with his own observations, stating that "AI is speeding up our research, but it's hard to say exactly how much."This accelerated progress comes as Anthropic CEO Dario Amodei has publicly advocated for a slower pace of AI development to allow safety measures to catch up. In a September essay, Amodei urged the AI industry to "slow the pace at which we improve the capabilities of AI models" due to concerns about AI's increasing ability to build subsequent generations of AI. The METR report's findings sharpen the conflict between these calls for restraint and the observed acceleration within Anthropic.The assessment measured model performance using Anthropic's internal Epoch Capabilities Index (AECI), a composite of benchmark scores. Since October 2025, AECI scores for Anthropic's models have grown at an estimated 1.9 times the rate observed between March 2024 and October 2025, with a 95% confidence interval ranging from 1.5 to 2.8 times. This implies that the company achieved roughly two years of historical progress in the last year.METR cautioned that while its highest-level conclusions about risk could be shared, Anthropic redacted a majority of the evidence considered most informative for these conclusions, citing commercially sensitive material. Despite these redactions, the report emphasizes the potential for further capabilities progress to lead to even faster acceleration, underscoring the importance of understanding and managing the risks associated with rapid AI advancement. Anthropic plans to release its own analysis of AI progress and safety practices in the coming weeks.