METR capability acceleration estimates range 0.99x–1.90x depending on methodology
tobyordoxford · x · 2026-11-10
Toby Ord continues the discussion with @austinc3301, endorsing the 1.6x estimate from all data points and sharing a METR sensitivity chart.
- The chart varies both the score calibration (AECI / system card / Epoch ECI) and the trend-estimation method (internal release dates, public dates, mixed internal+public)
- Acceleration estimates swing widely across combinations: 1.09x for all models with public dates, 1.90x for Anthropic-only with internal dates; one combination bottoms out at 0.99x
- Ord stresses there is still a lot of uncertainty about where the data points themselves should sit
Related event: Debate over METR capability acceleration: full-data fit yields 1.6x(3 posts)→
More from Models
- Fable 5.2 Reportedly Doubles the Best Remote Labor Index Score — we_are_mammals · 2026-11-10
- Musk says Grok 5 should surpass Claude Opus 5.5 within 2–3 months — XFreeze · 2026-11-10
- XBOW evals: GPT-6 vs Claude Mythos 5.1 exploit success per dollar on 65 vulnerable apps — mboehme_ · 2026-11-10
- Independent agent audit finds no evidence Fable stole the 3D aircraft model — victormustar · 2026-11-10
- Fable 5.2 works all weekend unattended: video editor built in 3 days — every · 2026-11-10
- Verity, a Local-Only AI Companion Mod for Minecraft via Ollama, Goes Viral — arnie_hacker · 2026-11-10