Overview
- The METR report, published November 5, 2026, estimates Anthropic’s benchmark capability growth has accelerated about 1.9 times since October 2025 with a 95% confidence interval of 1.5x–2.8x.
- METR attributes roughly 80% of that acceleration—a figure the group calls uncertain—to AI systems helping engineers with research and development rather than to more training compute.
- The review says Anthropic completed multiple unreleased internal models during the assessment window and identifies a ‘Model 4’ that cleared over 50% on every benchmark METR tracked.
- METR ran an 11-week embedded assessment with unprecedented internal access but notes substantial redactions, benchmarking limits, and weak evidence such as self‑reported Slack polls for productivity gains.
- Anthropic has said it will publish its own analysis of development pace and Responsible Scaling Policy compliance in the coming weeks, and the findings are likely to increase investor, regulator, and safety‑research scrutiny ahead of the company’s planned IPO.