Claude Fable 5

编程7 项

• ProgramBenchVals AI2.0%+
• Vibe Code Bench v1.1Vals AI90.4%+
• Terminal-Bench 4.0Vals AI41.4%+
• Terminal-Bench 4.0Artificial Analysis42.4%+
• Code MigrationVals AI55.1%+
• LiveBench CodingLiveBench86.0%+
• SciCodeArtificial Analysis61.0%+

科研7 项

• FrontierMath T1–3Epoch AI87.0%+
• Mystery Game PuzzlesEpoch AI52.0%+
• Humanity's Last ExamArtificial Analysis55.5%+
• CritPtArtificial Analysis28.6%+
• Terminal-Bench ScienceVals AI15.7%+
• Chess PuzzlesEpoch AI47.0%+
• LiveBench ReasoningLiveBench89.7%+

专业办公9 项

• APEX-AgentsMercor63.6%+
• Harvey 法律 Agent 基准Vals AI11.3%+
• AA-BriefcaseArtificial Analysis1,537.7+
• MedCodeVals AI56.1%+
• AA-Analyst AgentArtificial Analysis48.8%+
• LiveBench Data AnalysisLiveBench80.5%+
• GDPval-AAArtificial Analysis1,608.9+
• Finance AgentVals AI56.3%+
• AutomationBench-AAArtificial Analysis54.1%+

知识问答7 项

• LiveBench LanguageLiveBench90.7%+
• SimpleQA VerifiedEpoch AI70.7%+
• AA-OmniscienceArtificial Analysis43.3+
• LiveBench IFLiveBench75.8%+
• MLCR-AAArtificial Analysis20.6%+
• PBullshitBench v2BullshitBench56.0%+
• AA-LCRArtificial Analysis82.3%+

真人盲选1 项

• Arena TextLMArena1,503.3+

尚无成绩的评测

这些评测尚无可采用的公开成绩，缺测不补分。