Anthropic launched its Claude Fable 5.1 model on Tuesday, touting stronger coding, knowledge, and problem-solving performance while being cheaper for many users.
The company also rolled out Claude Mythos 5.1 to vetted users, which is essentially the same model but with more permissive safeguards in place for work in fields like cybersecurity.
Anthropic said Fable 5.1 outperformed Fable 5 as well as its flagship Opus 5 model across several benchmark tests. Before the latest rollouts, Opus 5 ranked as one of the strongest models on the market, though it is also among the most expensive, according to The Latent's 2026 index.
Fable 5.1 scored 55.8% on Terminal-Bench 4.0, a test of how well AI agents handle coding tasks. This is compared with a 42% score for Fable 5 and 52.3% for Opus 5. It also outperformed both models on tests covering scientific research, computer use, and broad reasoning.
Anthropic said Fable 5.1 should cost around 25% less than Fable 5 for most workloads and as much as 45% less for more complex tasks. These savings primarily stem from cutting the price of "cache reads," which allow the model to reuse information that it already processed.
Fable 5.1's outputs will also include an invisible AI watermark that Anthropic announced last month would be embedded into all new Claude models to comply with the EU's AI Act, which "requires AI providers serving its market to mark AI-generated content."
The move caused an uproar from users concerned that work they create using AI would be flagged, even for minor assistance in summarizing or rewriting a sentence. This prompted developers to start finding workarounds, with some already releasing tools that claim to weaken or scrub the watermark, though their efficacy is still unverified.