Claude Haiku 5.5

编程64.217 项

• ProgramBenchVals AI2.5%折算分 至多 80.4+
• MirrorCodeEpoch AI78.9%折算分 80.1+
• Vibe Code Bench v1.1Vals AI90.4%折算分 70.5+
• FrontierCode 2.0Artificial Analysis34.9%折算分 70.4+
• PFrontierSWE v2Proximal Labs47.4%折算分 70.3+
• CyberBench Patch v1.1Vals AI89.3%折算分 68.6+
• CFrontierCode 2.0Cognition27.7%折算分 67.2+
• APEX-SWEMercor55.1%折算分 67.1+
• Terminal-Bench 4.0Vals AI36.4%折算分 66.1+
• Terminal-Bench 4.0Artificial Analysis32.8%折算分 64.8+
• TapTap MakerTapTap Maker77.2%折算分 63.2+
• Code Migration v1.1Vals AI52.4%折算分 62.4+
• LiveBench Agentic CodingLiveBench57.2%折算分 61.1+
• HWeirdML v3Håvard Tveit Ihle9.8%折算分 58.7+
• LiveBench CodingLiveBench77.9%折算分 57.9+
• SciCodeArtificial Analysis55.0%折算分 56.4+
• Vibe Code Bench 1–100Vals AI20.4%+

科研63.313 项

• Chess PuzzlesEpoch AI55.0%折算分 69.9+
• MysteryMechanismVals AI35.1%折算分 69.4+
• Terminal-Bench ScienceArtificial Analysis20.0%折算分 67.1+
• CritPt v2 HardArtificial Analysis43.5%折算分 65.9+
• BioMysteryBenchVals AI70.4%折算分 65.5+
• FrontierMath T1–3Epoch AI75.1%折算分 64.7+
• CritPtArtificial Analysis18.9%折算分 63.9+
• Terminal-Bench ScienceVals AI10.0%折算分 62.3+
• Humanity's Last ExamArtificial Analysis44.4%折算分 61.5+
• Mystery Game PuzzlesEpoch AI30.0%折算分 59.8+
• AARC-AGI-2ARC Prize61.3%折算分 59.7+
• LiveBench ReasoningLiveBench84.6%折算分 58.8+
• 致知 · 逻辑月榜llm201432.3%折算分 57.9+

专业办公63.514 项

• AA-Analyst AgentArtificial Analysis53.8%折算分 73.3+
• AA-BriefcaseArtificial Analysis1,574.1折算分 70.8+
• Harvey LAB-AA v1.1Artificial Analysis0.6%折算分 至多 70.3+
• Finance Agent v2.1Vals AI65.4%折算分 68.0+
• GDPval-AAArtificial Analysis1,630.8折算分 67.6+
• Legal Research BenchVals AI43.8%折算分 66.9+
• TaxAgentBenchVals AI31.9%折算分 66.3+
• τ³-BankingMercor43.0%折算分 66.2+
• APEX-AgentsMercor56.1%折算分 64.5+
• MedCodeVals AI51.5%折算分 63.5+
• Excel Modeling v1.1Vals AI65.5%折算分 61.8+
• AutomationBench-AAArtificial Analysis35.4%折算分 55.4+
• Harvey 法律 Agent 基准 v1.1Vals AI4.2%折算分 至多 46.9+
• LiveBench Data AnalysisLiveBench51.8%折算分 36.6+

知识问答56.58 项

• MLCR-AAArtificial Analysis35.0%折算分 81.0+
• PBullshitBench v2BullshitBench57.0%折算分 65.6+
• AA-Omniscience v1.1Artificial Analysis12.3折算分 58.6+
• AA-LCRArtificial Analysis82.7%折算分 57.5+
• LNYT Connections 扩展版Lech Mazur65.7%折算分 51.1+
• LiveBench IFLiveBench54.7%折算分 44.0+
• LiveBench LanguageLiveBench62.6%折算分 41.6+
• SimpleQA VerifiedEpoch AI23.8%折算分 34.7+

真人盲选57.32 项

• Arena WebDevLMArena1,588.8折算分 61.2+
• Arena TextLMArena1,424.8折算分 40.6+

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。