METR estimates AI has nearly doubled the pace of Anthropic's AI research

What happened

METR reports that benchmark progress accelerated 1.9x since October 2025, with a 95% confidence interval of 1.5x to 2.8x.

METR attributes most of the acceleration to AI being used in AI research and development, rather than to scaling up compute. In other words, the gain comes mainly from AI helping to build AI, not from simply adding more computing power.

METR also sets limits on how the figure should be read. It says its estimates cannot be converted directly into company-wide R&D acceleration, so the 1.9x figure describes benchmark progress, not Anthropic's research as a whole.

Reports say Anthropic redacted much of the evidence METR found most informative.

Why it matters for creators

Faster lab research cycles suggest that model upgrades, including creative models, will keep arriving at a quicker pace.

The finding comes from one evaluation of one lab, with a wide confidence interval and redacted evidence, so it is best read as an indicator rather than a precise measure. Still, it is a rare outside estimate of how much AI tools are speeding up the labs that build them.