Frontier Log

AnthropicModel

Claude Fable 5.1

Made by
Anthropic
Released
Announced in
Introducing Claude Fable 5.1 and Claude Mythos 5.1anthropic.com
Event
Claude Fable 5.1 and Mythos 5.1
API record
/api/v1/models/claude-fable-5-1

Benchmark results

BenchmarkScoreSource check
Humanity's Last ExamNo tools60.9%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 2 other lab documents: 1 2
Humanity's Last ExamWith tools65.0%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 2 other lab documents: 1 2
reported later, in Introducing Claude Opus 5.565.6%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
SWE-bench Proaverage of 5 trials · The source doesn't say whether this is the public or the commercial set81.2%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
ARC-AGI-2max effort; semi-private set; as reported by the ARC Prize Foundation90%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 1 other lab document: 1
Terminal-Bench 4.055.8%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 5 other lab documents, the first 3 linked: 1 2 3
Terminal-Bench Science 0.1standard error ±3.5 to 4.5 points52.6%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
AutomationBenchZapier leaderboard, private held-out set; max effort; with production safeguards31.4%
CorroboratedThe same number is printed in 3 other lab documents: 1 2 3
OSWorld 2.1Full set, partial score · max effort; mean of 5 runs; September 10, 2026 task files; the system card labels it OSWorld 2.0 · reported later, in Introducing Claude Opus 5.580.7%
CorroboratedThe same number is printed in 1 other lab document: 1
OSWorld 2.0Offline set, partial score · partial score; August 2026 task release, not comparable to earlier OSWorld 2.0 scores; tasks where safeguards intervened scored zero77.9%
CheckedA person checked this number against its source on 18 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
FrontierCode 1.1Main set · medium effort50.9%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
run by Cognition in Claude Code · reported later, in Introducing Claude Opus 5.550.3%
CorroboratedThe same number is printed in 1 other lab document: 1
best effort level; run by Cognition in Claude Code · reported later, in Claude Opus 5.5 System Card52.8%
CheckedA person checked this number against its source on 20 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
FrontierCode 1.1Extended set · medium effort63.6%
CheckedA person checked this number against its source on 19 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
DeepSWE v1.1average of 5 trials67.4%
CheckedA person checked this number against its source on 17 Oct 2026. This confirms the number matches what the lab published, not that the score was reproduced.
CorroboratedThe same number is printed in 2 other lab documents: 1 2
Sources

Reported by Anthropic

Events

Stories

All models