Claude Fable 5.2 (Anthropic)

Anthropic released Claude Fable 5.2 on November 9, 2026 as a faster and cheaper version of its most powerful model, pricing it at $8 per million input tokens and $40 per million output tokens on the API, about 20% below Fable 5.1, with writes up to 30% faster. The model keeps a 1M-token context window and 128,000 maximum output tokens, and the API model ID is claude-fable-5-2, with optional fast mode running up to 2.5 times faster at $16 and $80 per million. MadRobot notes Anthropic's published figures show Fable 5.2 ahead of Fable 5.1 on every test, including 49.1% on FrontierCode 2.0 and 69.9% on FrontierSWE v2, though Gemini 4 Carbon beat it on AutomationBench and Opus 5.5 edged it on one knowledge work rating. The figures have not been independently checked. One engineer reportedly had the model coordinate over 1,000 subagents for eight days to produce a prototype operating system that boots on a laptop.

Anthropic's product page lists Claude Fable 5.2 as the current generation for the hardest knowledge work and coding problems, available to Pro, Max, Ultra, Team, and Enterprise users and to developers on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Foundry. Developers are directed to the model ID claude-fable-5-2 through the Claude API. List pricing is $8 per million input tokens and $40 per million output tokens, down from $10 and $50 for Fable 5.1, with cache reads at $0.20 per million tokens. Fast mode runs on the Claude Platform, Google Cloud under both management options, Microsoft Foundry, and Claude Code, delivering up to 2.5x faster speed at $16 input and $80 output per million tokens, and US-only inference is offered at 1.1x pricing.

Anthropic's system card for Claude Fable 5.2 and Claude Mythos 5.2, dated November 9, 2026, describes the two names as two configurations of one new large language model. Fable 5.2 is released for general use with added safeguards blocking certain tasks in high-risk dual-use domains such as biology and cybersecurity, while Mythos 5.2 has more permissive safeguards and is restricted to vetted individuals and organizations through trusted access programs. The card reports pre-deployment evaluations across seven areas, stating that the models advance the frontier in coding, knowledge work, and problem-solving with better novel mathematical and scientific reasoning. On the RSP autonomy threat model for high-stakes misalignment, Mythos 5.2 is judged to remain at the low risk level from Anthropic's August 2026 Risk Report, and on chemical and biological risks it is a modest improvement over Opus 5.5 and Mythos 5.1 without crossing the CB-2 threshold.

Claude Fable 5.2 is documented for demanding reasoning and long-horizon agentic work, such as multi-hour agent sessions, multistep deep research, and analysis delivered as a finished document, spreadsheet, or deck. The spec sheet lists a 1M-token context window, 128K maximum output, adaptive thinking that is always on, a June 2026 knowledge and training cutoff, moderate comparative latency, and a release date of November 9, 2026. The API model ID is claude-fable-5-2, with matching identifiers on Amazon Bedrock, Microsoft Foundry, Claude Platform on AWS, and Google Cloud under both Google- and Anthropic-managed offerings. Pricing is $8 input and $40 output per MTok, with cache writes at $10 (5m) and $16 (1h), cache reads at $0.20, and a 50% Batch API discount. Forced tool choice types and sampling parameters return 400 errors, and retirement is not sooner than November 9, 2027.

Anthropic announced Claude Fable 5.2 and Claude Mythos 5.2 on November 9, 2026, describing improvements in coding, reasoning, and vision, with outputs generated up to 30% faster than the Fable 5.1 and Mythos 5.1 generations. Standard pricing is $8 per million input tokens and $40 per million output tokens, with an optional fast mode at $16 and $80 per million that speeds token generation by up to 2.5 times. The two names are the same underlying model, with Fable carrying stronger cybersecurity and biology safeguards that let it ship to all users, while Mythos requires access through Anthropic's expanded Life Sciences and Cyber Verification programs. Anthropic says prerelease testers found the models communicate more clearly and used them for software migrations, electronics design, black-hole physics problems, robotics evaluation work, and a 30-hour formalization of Fermat's Last Theorem proof.