Introducing Claude Fable 5.1 and Claude Mythos 5.1 Anthropic

• Fable 5.1
• Fable 5

Terminal-Bench-Science 0.1: The standard error is ±3.5–4.5 pts per model. The public leaderboard (3 trials/task, Claude Code harness) reports Claude Opus 5 at 30.0% and Claude Fable 5 at 21.4%; our setup reproduces them at 29.0% and 24.7%, respectively, both within noise

• Mythos 5.1
• Fable 5.1
• Mythos 5

Terminal-Bench 4.0 scores by cost (log scale), at each effort level. Claude Fable 5.1 and Claude Mythos 5.1 are the same underlying model; the gap between them reflects the tasks on which our earlier, less precise cyber safeguards intervened. With the improvements we’re making to these safeguards today, we expect the difference between the models to be much smaller

• Fable 5.1 (with tools)
• Fable 5.1 (no tools)
• Fable 5 (with tools)
• Fable 5 (no tools)

Humanity’s Last Exam scores by cost (log scale), at each effort level. CursorBench 3.2.0 scores by cost (log scale), at each effort level

• Fable 5.1
• Fable 5

CursorBench 3.2.0 by cost (log scale), at each effort level

“In internal benchmarks, Claude Fable 5.1 solves more of our coding problems than Fable 5 or Opus 5, and achieves state of the art on trading intuition. While prior models became hard to follow the longer they worked, Fable 5.1 remains readable over long, multi-step tasks.”

CompanyJane Street Capital
AuthorCraig Falls, Head of Quantitative Research
01 / 22

Inference speedup for seven open-

• Original implementation
• Optimized

Estimated GPU cost of three genome-wide analyses before and after optimization, at cloud list price. Evo 2 40B saves more on a whole job (2.3x) than per forward (1.4x) because some of its optimizations only pay off across many sequences

• Cache reads
• All other tokens

Indexed cost of running the same workloads on Fable 5 and Fable 5.1, at usage-based pricing measured at default effort over four weeks of actual usage in August 2026. Typical workload covers Fable usage across Claude Enterprise, Claude Code, and the API. Highly agentic workload covers context-heavy, tool-heavy work, where cache reads make up most of the cost

Try ClaudeStart building

1Terminal-Bench-Science 0.1: The standard error is ±3.5–4.5 pts per model. The public leaderboard (3 trials/task, Claude Code harness) reports Claude Opus 5 at 30.0% and Claude Fable 5 at 21.4%; our setup reproduces them at 29.0% and 24.7%, respectively, both within noise

²OSWorld 2.0: Scores are on the benchmark authors’ August 2026 task release; Fable 5 and Opus 5 were re-run under the same conditions. Because the task files differ from earlier releases, these numbers aren’t directly comparable to previously published OSWorld 2.0 results, which is why no competitor score is shown

3These three targets are (EGFR, Nipah G, 15-PGDH) and come from Adaptyv Bio’s protein design competitions. The Nipah G comparison is against de novo designs targeting the receptor-binding site on the G head (best: ~8–12 nM, N1032). A de novo entry from Nick Boyd/Escalante Bio that targets a different region (the stalk) reached ~1.4 nM (design_7), comparable to our best binder

Source: www.anthropic.com