Anthropic released Claude Fable 5.2 on November 9, 2026, positioning the large language model as its most capable offering for complex reasoning, agentic coding, and long-horizon technical work. Artificial Analysis benchmarks released the same day show Fable 5.2 variants immediately claiming the top HLE-Diamond scores at 80.2% (max effort), 75.8% (xhigh), and 73.8% (high), well ahead of Gemini 4 Carbon’s prior leading mark of 69.1%. Gains appear strongest in the reasoning partition, with solid knowledge performance, reflecting improvements over Fable 5.1 and Opus 5.5 on related evaluations like FrontierCode and Terminal-Bench-Science. Traders are pricing in a strong debut because the model’s verified results on this refined 1,000-question benchmark set a new ceiling, though future independent runs or effort-level variations could still shift exact thresholds before market resolution.