METR assessment of Anthropic R&D acceleration
2026-11-05 — headlined Anthropic R&D acceleration — METR published a high-access independent assessment of Anthropic (evaluation period August 12–October 29, 2026) finding Anthropic is likely improving model capabilities at about 2x its March 2024–October 2025 rate (AECI growth estimated at 1.9x since October 2025, CI 1.5–2.8x — roughly two years of historical progress in the last year), with a majority of the acceleration likely attributable to Anthropic's own use of AI in AI research rather than compute scaling; METR found no significant bottlenecks over the next 12 months and warns unimpeded capability growth could accelerate further. Anthropic redacted a majority of the most informative evidence, described as protecting other labs from acceleration. A companion METR review of the public August 2026 Risk Report's "Risks from automated R&D" section agrees catastrophic risk from the July 15 models was low but says the public report "understates the level of acceleration" and that Anthropic's 2x-acceleration early-warning threshold plausibly had been crossed (METR · METR · The Information). Watch: Anthropic's response, comparable assessments at other labs, and whether the 2x rate estimate holds as newer data lands.