Claude Fable 5.2

Anthropic’s November 9 launch brought Claude Fable 5.2 and Mythos 5.2 to paid Claude plans and API customers, with Mythos limited to vetted organizations. Both variants share a 1M-token context window, support up to 128K output tokens, and use adaptive thinking that remains continuously enabled. The standard API price fell to $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for the 5.1 models. Anthropic says outputs are up to 30% faster, while an optional fast mode can generate tokens up to 2.5 times faster at twice the standard price.

Anthropic's November 9, 2026 system card introduces Claude Fable 5.2 and Claude Mythos 5.2 as two configurations of a new large language model, with Fable intended for general use and Mythos limited to vetted individuals and organizations through trusted access programs. The card reports that the models advance the frontier in coding, knowledge work, and problem-solving. Pre-deployment evaluations covered seven areas, including Responsible Scaling Policy and Frontier Compliance Framework checks, with the autonomy threat model rated at a low risk level consistent with Anthropic's August 2026 Risk Report. On chemical and biological risks, the more permissive configuration was only a modest improvement over Claude Opus 5.5 and Mythos 5.1 and did not cross the CB-2 threshold, so CB-1 protections were applied.

Anthropic's model documentation describes Claude Fable 5.2 as built for demanding reasoning and long-horizon agentic work, including agent sessions running for hours, multistep deep research, and analysis delivered as a finished document, spreadsheet, or deck. Anthropic lists the model ID claude-fable-5-2 for Claude API, Amazon Bedrock, Microsoft Foundry, and both Google Cloud variants. The model offers a 1M token context window and 128K max output, with adaptive thinking always on and controlled through the effort parameter; temperature, top_p, and top_k must be omitted, and forced tool choice returns a 400 error. Pricing is $8 per MTok input and $40 per MTok output, with $0.20 cache reads, and both knowledge and training cutoffs sit at June 2026.

The LLM Stats leaderboard places Claude Fable 5.2 first among 307 coding models, first among 407 reasoning models, first among 244 vision models, second among 353 math models, and second among 233 tool-calling models. These rankings summarize benchmark data sourced from Anthropic materials rather than independent measurements. The listed benchmark results include a 97% score on SWE-bench Multilingual and 1,831 Elo on GDPval-AA v2.1 at maximum effort. The page also reports a blended price of $9.52 per million tokens and an overall LLM Stats Score of 61.3, but its data remains a re-index of Anthropic’s published results.

Anthropic introduced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, describing the models as designed for demanding technical work. The release reports improvements in coding, reasoning, and vision, along with clearer communication and output generated up to 30% faster than the 5.1 models. Both names represent the same underlying model, but Fable includes cybersecurity and biology safeguards for broad availability, while Mythos is restricted to approved Life Sciences and Cyber Verification Programs. Standard API pricing is $8 per million input tokens and $40 per million output tokens, while optional fast mode can accelerate token generation by up to 2.5 times.