Gemini 3.1 Pro Preview

编程8 项

• ProgramBenchVals AI0.0%+
• SciCodeArtificial Analysis58.7%+
• LiveBench CodingLiveBench76.5%+
• Terminal-Bench 4.0Artificial Analysis4.0%+
• Code MigrationVals AI17.3%+
• Vibe Code Bench v1.1Vals AI32.0%+
• Vibe Code Bench 1–100Vals AI6.7%+
• Terminal-Bench 4.0Vals AI2.5%+

科研8 项

• Chess PuzzlesEpoch AI51.0%+
• Humanity's Last ExamArtificial Analysis47.0%+
• CritPtArtificial Analysis17.7%+
• Mystery Game PuzzlesEpoch AI34.0%+
• MysteryMechanismVals AI20.7%+
• LiveBench ReasoningLiveBench84.0%+
• FrontierMath T1–3Epoch AI59.6%+
• Terminal-Bench ScienceVals AI1.4%+

专业办公9 项

• MedCodeVals AI59.1%+
• LiveBench Data AnalysisLiveBench78.5%+
• AA-Analyst AgentArtificial Analysis41.3%+
• Harvey 法律 Agent 基准Vals AI0.0%+
• AutomationBench-AAArtificial Analysis35.4%+
• Finance AgentVals AI43.0%+
• APEX-AgentsMercor35.1%+
• GDPval-AAArtificial Analysis789+
• AA-BriefcaseArtificial Analysis456.4+

知识问答7 项

• SimpleQA VerifiedEpoch AI73.5%+
• LiveBench IFLiveBench79.1%+
• AA-OmniscienceArtificial Analysis31.9+
• LiveBench LanguageLiveBench85.4%+
• AA-LCRArtificial Analysis82.0%+
• MLCR-AAArtificial Analysis11.1%+
• PBullshitBench v2BullshitBench31.0%+

真人盲选1 项

• Arena TextLMArena1,487+

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。