Key facts
- METR's 11-week assessment found Anthropic's benchmark capabilities improved 1.9 times faster since October 2025 than previously.
- METR estimated AI accounted for perhaps 80% of the acceleration, a figure it said was uncertain.
- METR said it found no convincing evidence of major bottlenecks that could slow progress over the next year.
- METR cautioned AI was not yet capable of highly significant, fully autonomous contributions to the most important stages of model training.
- METR said the acceleration was not due to an acceleration in the scaling of training compute.
An independent study by the AI safety nonprofit METR found that Anthropic's AI models are improving nearly twice as fast as before, as the technology plays a bigger role in developing more powerful versions of itself. The findings were published on Thursday, November 5, 2026, according to Reuters.
METR, also known as Model Evaluation and Threat Research, conducted an 11-week assessment drawing on confidential research, interviews and performance data from unreleased models. It found Anthropic's benchmark capabilities had been improving 1.9 times faster since October 2025 than previously.
METR estimated that AI accounted for perhaps 80% of the current acceleration, though it cautioned that the figure was uncertain. The researchers said they found no convincing evidence of major bottlenecks that could slow progress over the next year.
"We are confident that the acceleration was not due to an acceleration in the scaling of training compute," METR said. The nonprofit also cautioned that AI was not yet capable of making highly significant, fully autonomous contributions to the most important stages of model training.
The findings offer some of the clearest evidence yet that AI is accelerating its own development, potentially paving the way for recursive self-improvement, when AI systems can build more powerful versions of themselves without human intervention. That prospect has fueled warnings from AI researchers that such systems could become difficult to align with human goals, monitor and control.