Compare / Claude Fable 5.1 vs Claude Opus 4.7

Claude Fable 5.1 vs Claude Opus 4.7

22 results from 7 sources that measured both models. Each row is in the source's own units; there is no overall winner.

Claude Fable 5.1
Anthropic · profile
Claude Opus 4.7
Anthropic · profile
Results compared
22
Shared sources
7
BenchmarkMetricClaude Fable 5.1Claude Opus 4.7
APEX-AgentsreportedTasks passed% of tasks 68.6%49.2%
Arena (formerly LMArena)reportedOverallArena rating 1,4931,494
Arena (formerly LMArena)reportedOverallArena rating 1,5011,477
Arena (formerly LMArena)reportedBusiness, management and financeArena rating 1,4851,508
Arena (formerly LMArena)reportedCreative writingArena rating 1,4841,488
Arena (formerly LMArena)reportedExpert promptsArena rating 1,5161,535
Arena (formerly LMArena)reportedInstruction followingArena rating 1,4971,503
Arena (formerly LMArena)reportedOverallArena rating 1,5011,502
Arena (formerly LMArena)reportedWriting, literature and languageArena rating 1,4961,496
Artificial AnalysisreportedArtificial Analysis Coding Indexindex score 81.673.6
Artificial AnalysisreportedArtificial Analysis Intelligence Indexindex score 53.440.7
Artificial AnalysisreportedGPQA Diamond% of questions 93.7%91.4%
Artificial AnalysisreportedHumanity's Last Exam% of questions 59.1%42.3%
Artificial AnalysisreportedLong-context reasoning (AA-LCR)% of questions 85.3%78.7%
Artificial AnalysisreportedTerminal-Bench 2.1% of tasks 91.4%83.1%
Artificial AnalysisreportedΤ-bench banking% of tasks 47.2%34.6%
GDP.pdfreportedProfessional document tasks% of rubric 29.6%21.0%
SimpleQA Verified (Epoch AI)reportedCorrect answers% of questions 70.8%51.7%
UGI LeaderboardreportedRequested-length error% off the requested word count 3.0%1.0%
UGI LeaderboardreportedStyle adherencescore from 0 to 1 0.430.41
UGI LeaderboardreportedWriting scorescore out of 100 63.961.4
Vending-Bench 2reportedMoney after a yearUS dollars $5,421.56$10,936.76

Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.