Redwood and Anthropic Launch Conceptual Reasoning Index to Evaluate AI Safety Reasoning

RyanGreenblatt · x · 2026-08-13

Redwood AI, in collaboration with Anthropic, developed the Conceptual Reasoning Index (CRI) to evaluate AI performance on safety-related reasoning. Every data point is manually checked by researchers. The chart shows the highest-scoring models from each company plus Fable 5, Muse Spark 1.2, and Gemini Flash 3.6.

Related event: Redwood and Anthropic Introduce Conceptual Reasoning Index(2 posts)→

Original post →

More from Safety

Safety channel →