Claude Fable 5.1 sets new standard for coding and scientific research

On September 1, 2026, Anthropic announced the release of Claude Fable (and Mythos) 5.1, according to a report published the same day by Simon Willison. Anthropic’s announcement states that the new model “sets a new standard for coding, knowledge work, and long-running problem-solving tasks.”

The company’s release materials devote considerable attention to scientific research performance. Anthropic cites a score of 52.6% on Terminal-Bench-Science 0.1, a benchmark described as newly introduced. The benchmark itself was first announced separately, though the source material available here does not specify the announcing party or date beyond a partial reference.

Willison’s post, titled “Claude Fable 5.1 made me a really nice animated pelican,” frames the release within his ongoing practice of testing new models against a standard illustrative task. The available summary does not include further detail on the pelican test’s results, nor does it elaborate on Fable 5.1’s coding benchmarks beyond the scientific research figures cited above.

No additional named sources, dates, or benchmark comparisons are present in the material provided. The Specola notes that Anthropic’s characterization of the release as setting a “new standard” is presented here as reported company language, not as an independently verified claim.