Peter Gostev conducted a comparative analysis of Anthropic's Claude Fable 5.2 and Opus models. The evaluation involved over 100 generations across various creative tasks to assess performance consistency.
The Remote Labor Index benchmark showed Fable 5.2 achieving a score of 41.3%, compared to 21.3% for Opus 5.5. Initial benchmarks suggest the new models operate up to 30% faster than previous versions.
Anthropic released the Fable 5.2 and Mythos 5.2 models on November 9, 2026. These models are designed for tasks involving coding, reasoning, vision, and long-term decision-making.
Standard pricing for Fable 5.2 is set at $8 per million input tokens and $40 per million output tokens. A fast mode is available at double these rates, costing $16 and $80 respectively.
Access to the Mythos 5.2 model is currently restricted to participants in Anthropic’s Life Sciences and Cyber Verification Programs. The Fable line was originally introduced in June 2026.
