Claude Fable 5.2

What it is and who made it

Anthropic, an AI company, released Claude Fable 5.2 on November 9, 2026. It is part of the Claude series of reasoning models — AIs built to think through multi-step problems before answering. Fable 5.2 shares its base with Claude Mythos 5.2, but Fable has stricter safety guards, especially around cybersecurity and biology, making it safe for general use. Earlier industry estimates placed Claude Fable at around 5 trillion parameters, though these figures may not apply specifically to version 5.2.

Why it matters and how it performs

The update makes the model up to 30 percent faster at generating responses and lowers prices to $8 per million input tokens and $40 per million output tokens (fast mode costs double). On benchmark tests, Fable 5.2 scores above Claude Opus 5.5 in nearly every category, including high marks for coding, software engineering, and cybersecurity tasks. It also received the highest Intelligence Index score (65 out of 100) on a widely tracked leaderboard. However, it is expensive compared to other models and slower, taking about 44.6 minutes per task on average versus 12.7 minutes for similar systems.

What you can do with it

With a context window of one million tokens (roughly 1,500 A4 pages), users can feed massive documents into the model. Testers have used it to complete difficult software migrations, design three working electronic gadgets, solve two open problems in black hole physics, improve robotics assessments, and even formalize a proof of Fermat’s Last Theorem in 30 hours. The model supports both text and image input. For easier command-line access, developers can use the free open-source tool llm-anthropic, updated to version 0.31. Note that using the highest reasoning setting ("max") can sometimes cause the model to run out of tokens and crash.

Questions people ask

How much does Claude Fable 5.2 cost?

Standard pricing is $8 per million input tokens and $40 per million output tokens; an optional fast mode doubles those rates. It is priced higher than many competitors.

How well does it perform compared to other AIs?

On independent benchmarks, it scores above Claude Opus 5.5 in nearly every category and earned the highest Intelligence Index score (65/100) recorded so far, though it runs slower and uses more tokens to reach answers.

Can I try it without clicking around a website?

Yes, a free open‑source tool called llm‑anthropic (version 0.31) lets you talk to the model directly from your computer's command line by typing prompts.

Written by AI only from the stories listed below, each of which links to its own sources.