Released Claude Fable 5.2 and Claude Mythos 5.2, with improved performance for coding, reasoning, and agent tasks; access to Mythos is restricted to approved organizations. (Nov. 10)γ10γ
Anthropic launched Claude Fable 5.2 and access-restricted Claude Mythos 5.2 on Nov. 9, 2026, for coding, reasoning and long-running agent work. The models share an underlying system; Mythos is reserved for organizations vetted through Anthropicβs cyber and life-sciences programs, while Fable is generally available. Both have a 1-million-token context window and 128,000-token output limit. Anthropic says they generate outputs up to 30% faster than version 5.1; input pricing fell from $10 to $8 per million tokens and output from $50 to $40. Optional fast mode costs $16/$80 per million input/output tokens. In company-reported tests, Fable 5.2 scored 69.9% on SWE v2, versus 56.3% for 5.1; Mythos 5.2 reached 89.6% on CVE rediscovery, versus Mythos 5.1βs 78.3%. The restricted rollout reflects the dual-use stakes of stronger cyber capabilities. Anthropic introduced Mythos in April and a safeguarded public Fable variant two months later; the 5.2 launch also intensifies competition with OpenAI and Google on price and technical performance.