Anthropic releases Claude Fable 5.2 for long-horizon technical work
The short version
Anthropic released Claude Fable 5.2, its active Fable model for demanding coding and long-horizon agent work, with a 1M-token context, 128K maximum output, and lower list pricing than Fable 5.1.
What changed
Claude Fable 5.2 replaces Fable 5.1 as the current model ID, adds always-on adaptive thinking and changes several API migration behaviors.
What it means for you
Teams using Fable-class workloads can evaluate a new model and lower $8/$40 MTok list price, but should test compatibility, effort settings, and task-level cost before changing production defaults.
Anthropic released Claude Fable 5.2 as its current Fable model for demanding reasoning, long-running agent sessions, and difficult technical work. The model is available through the Claude API and Anthropic lists deployment paths across its own products, AWS, Google Cloud, and Microsoft Foundry.
The official model documentation lists a one-million-token context window, a 128K maximum output, adaptive thinking that is always enabled, and $8/$40 per million input/output tokens. Fast mode is optional and costs more. The release also changes API behavior from Fable 5.1, including thinking-block compatibility and prefix-check handling.
This is a new model release rather than a routine client update. Teams should evaluate it on representative long-horizon tasks, recording completion quality, latency, total tokens, and fallback behavior before changing a production default. Anthropic recommends starting most workloads with its latest Opus model and moving to Fable when those evaluations remain insufficient.
Related tools
Fact check
Coverage timeline
- Primary sourceAnthropicIntroducing Claude Fable 5.2 and Claude Mythos 5.2
- Primary sourceAnthropicClaude Fable 5.2 overview