Claude Opus 5.5 vs Gemini 3.1 Pro
完整资料:Claude Opus 5.5(Anthropic)和 Gemini 3.1 Pro(Google),包括全部基准测试、价格和排名。
更新于 · 数据来自 14 个来源
截至 2026年11月16日,质量上 Claude Opus 5.5 领先 Gemini 3.1 Pro(BenchLeader Index 68.4 比 60.4)。不过,Gemini 3.1 Pro 更便宜,对方的价格是它的 1.8 倍(每百万 token 混合价格 $4.50 比 $8.00)。在两者都有成绩的所有类别中,Claude Opus 5.5 均领先。
要点(截至 2026年11月16日)
- Claude Opus 5.5 的 BenchLeader Index 为 68.4,Gemini 3.1 Pro 为 60.4(BenchLeader,2026年11月16日)。
- Claude Opus 5.5 每百万 token 混合价格为 $8.00(标价,2026年11月16日)。
- Gemini 3.1 Pro 每百万 token 混合价格为 $4.50(标价,2026年11月16日)。
- 在两者都有结果的 38 项基准测试中,Claude Opus 5.5 领先 34 项(BenchLeader,2026年11月16日)。
按用途怎么选
推荐模型的数值列在前面。类别得分采用 BenchLeader 标尺(50 为模型平均水平,15 分为一个标准差)。领先半个标准差及以上为“高”置信度;相差不到 1 分则视为“均可”。
| 用途 | 推荐 | 理由 | 置信度 |
|---|---|---|---|
| 编程 | Claude Opus 5.5 | 类别得分 79 比 59;IOI 95.1% 比 51.8% | 高 |
| 智能体与工具调用 | Claude Opus 5.5 | 类别得分 61 比 57;LiveBench Agentic Coding 71.7% 比 44.1% | 中 |
| 数学 | Claude Opus 5.5 | 类别得分 73 比 66;OTIS Mock AIME 100.0% 比 95.6% | 高 |
| 推理 | Claude Opus 5.5 | 类别得分 72 比 65;GPQA Diamond 90.6% 比 94.4% | 中 |
| 知识 | Claude Opus 5.5 | 类别得分 73 比 66;SimpleQA Verified 72.2% 比 73.5% | 中 |
| 成本最低 | Gemini 3.1 Pro | 每百万 token 混合价格 $4.50 比 $8.00;智能体编程会话 $0.168 比 $0.320 | – |
| 首次回答最快 | Gemini 3.1 Pro | 首个回答 token 15.2 秒(Gemini 3.1 Pro (high))比 380.2 秒(Claude Opus 5.5 (max)) | – |
| 低预算大批量 | 两者都不选 | 通常更小的模型更合适:便宜好用的大模型、最快的大模型 | – |
主要差异
- 质量上 Claude Opus 5.5 (max) 领先:68.4 比 60.4。
- Claude Opus 5.5 (max) 在智能体与工具、编程、知识、数学、推理方面更强。
- Gemini 3.1 Pro (high) 更便宜,对方的价格是它的 1.8 倍(每百万 token 混合价格 $4.50 比 $8.00)。
概览
| 规格 | Claude Opus 5.5 | Gemini 3.1 Pro |
|---|---|---|
| 提供商 | Anthropic | |
| 发布日期 | 2026年9月22日 | 2026年2月19日 |
| 上下文窗口(token) | 1,000,000 | 1,048,576 |
| 最大输出(token) | 128,000 (up to 300K on the Message Batches API (beta)) | – |
| 输入类型 | 文本、图像 | – |
| 推理强度设置 | low, medium, high, xhigh, max | – |
| 对比时的配置 | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) |
| 开放权重 | 否 | 否 |
| 每百万 token 价格(输入 / 输出 / 缓存输入) | $4.00 / $20.00 / $0.200 | $2.00 / $12.00 / $0.200 |
| API 模型 ID | claude-opus-5-5 | – |
| 使用渠道 | Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry | – |
上限、可用性和已公布的价格均来自厂商官方页面: platform.claude.com (2026年10月20日 核实)。其余内容均来自当日数据。
逐类别对比
类别得分采用 BenchLeader 标尺:50 为参评模型的平均水平,15 分为一个标准差。差距不到 1 分视为持平。
| 类别 | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) | 领先者 |
|---|---|---|---|
| 推理 | 72.1 | 64.7 | Claude Opus 5.5 |
| 编程 | 78.8 | 58.6 | Claude Opus 5.5 |
| 智能体与工具 | 61.3 | 56.7 | Claude Opus 5.5 |
| 数学 | 73.2 | 65.7 | Claude Opus 5.5 |
| 知识 | 72.5 | 65.6 | Claude Opus 5.5 |
| 多模态 | 70.2 | – | 无法比较 |
| 长上下文 | 67.1 | – | 无法比较 |
| 综合 | 70.2 | – | 无法比较 |
同一图表对比
各领域采用与 Index 相同的标尺,平均模型为 50 分。单击某个领域可查看构成它的基准测试。 可从列表中最多选择 5 个模型。
未绘制(测得的模型少于两个):指令遵循, 长上下文, 多模态, 人类偏好。
各来源的分歧
在两者都有结果的 38 项基准测试中,Claude Opus 5.5 领先 34 项,Gemini 3.1 Pro 领先 4 项。所有同时评测两者的综合评分都认为 Claude Opus 5.5 更强。
| 综合 Index | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) | 领先者 |
|---|---|---|---|
| BenchLeader Index | 68.4 | 60.4 | Claude Opus 5.5 |
| Arena Text | 1507† | 1487 | Claude Opus 5.5 |
各发布方使用各自的任务、评测框架和推理强度设置,因此总体结论可能不同;BenchLeader 将它们统一到同一标尺上。† 表示同一模型在另一推理强度设置下的测量值(发布方未测试本页对比的设置)。
价格、速度与延迟
标价按每 1 个输出 token 对应 3 个输入 token 混合计算。每项任务的价格采用方法论页面中的 token 数量。每项基准测试任务的成本、token 数和耗时均为实测值,取自 Artificial Analysis 整个 Intelligence Index 测试的平均值。
| 指标 | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) |
|---|---|---|
| 每百万 token 混合价格 | $8.00 | $4.50 |
| 每百万 token 输入 / 输出 | $4.00 / $20.00 | $2.00 / $12.00 |
| 每百万 token 缓存输入 | $0.200 | $0.200 |
| 每项任务价格:聊天回复 | $0.0076 | $0.0044 |
| 每项任务价格:总结一份 30 页的报告 | $0.060 | $0.031 |
| 每项任务价格:代码编辑 | $0.054 | $0.030 |
| 每项任务价格:智能体编程会话 | $0.320 | $0.168 |
| 每项任务价格:结构化信息提取 | $0.012 | $0.0064 |
| 智能体编程会话(输入大多命中缓存) | $0.149 | $0.087 |
| 每项基准测试任务成本(AA Index) | $10.04 | – |
| 每项任务输出 token(AA Index) | 173k | – |
| 其中推理部分 | 120k | – |
| 每项任务耗时(AA Index) | 25 min | – |
| 输出速度 | 116 tok/s | 111 tok/s |
| 首次回答时间 | 380.2 秒 | 15.2 秒 |
| 完整回答耗时(最近一次测量 2026年11月13日) | 711.4 秒 | 37.5 秒 |
每个模型都按其排名最高的推理强度设置进行对比。Claude Opus 5.5 (max) 得分 68.4,首个回答 token 需 380 秒;Claude Opus 5.5 (high) 为 66.0(26 秒),Claude Opus 5.5 (xhigh) 为 65.4(78 秒)。Gemini 3.1 Pro (high) 得分 60.4,首个回答 token 需 15 秒;Gemini 3.1 Pro 为 59.9(15 秒)。推理强度越高,回答前思考越久:得分更高,但等待更长、计费 token 也更多。 推理强度如何影响得分和成本
全部基准测试
共 221 项基准测试,按类别分组;点击名称可查看该排行榜及领先模型。† 表示同一模型在另一推理强度设置下的测量值,仅供参考,不计入胜场。
| 指标 | 发布方 | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 68.4 | 60.4 |
| 智能体与工具得分 | BenchLeader | 61.3 | 56.7 |
| 编程得分 | BenchLeader | 78.8 | 58.6 |
| 综合得分 | BenchLeader | 70.2 | – |
| 知识得分 | BenchLeader | 72.5 | 65.6 |
| 长上下文得分 | BenchLeader | 67.1 | – |
| 数学得分 | BenchLeader | 73.2 | 65.7 |
| 多模态得分 | BenchLeader | 70.2 | – |
| 推理得分 | BenchLeader | 72.1 | 64.7 |
| 混合价格(美元/百万 token) | BenchLeader | $8.00 | $4.50 |
| 输出速度 | BenchLeader | 116 tok/s | 111 tok/s |
| 首次回答时间 | BenchLeader | 380.2 秒 | 15.2 秒 |
| 上下文窗口 | BenchLeader | 1,000,000 token | 1,048,576 token |
| 推理 | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 90.6% | 94.4% |
| Humanity's Last Exam | Scale AI / CAIS | – | 46.4%† |
| SimpleBench | SimpleBench | 88.4%† | 79.6%† |
| Arena Hard Prompts | Arena (LMArena) | 1530† | 1508† |
| LiveBench 2.0 Reasoning | LiveBench | 92.2% | – |
| LiveBench 1.0 Reasoning | LiveBench | 92.2% | 84.0% |
| GPQA Diamond (AA) | Artificial Analysis | – | 94.1%† |
| Humanity's Last Exam (AA) | Artificial Analysis | 61.4% | 47.0%† |
| HLE-Diamond (AA) | Artificial Analysis | 67.9% | 31.4%† |
| GPQA Diamond (Vals) | Vals AI | – | 95.5% |
| EnigmaEval | Scale AI SEAL | – | 36.8% |
| ARC-AGI-1 | ARC Prize | 97.5% | 98.0%† |
| ARC-AGI-2 | ARC Prize | 91.7% | 77.1%† |
| ARC-AGI-3 | ARC Prize | – | 0.4%† |
| CritPt | Artificial Analysis | 31.7% | 17.7%† |
| CritPt v2 hard subset (AA) | Artificial Analysis | 68.0% | – |
| MysteryMechanism | Vals AI | 49.5% | 20.7% |
| HLE-Diamond (Scale) | Scale AI SEAL | 55.0%† | – |
| Arena Expert | Arena (LMArena) | 1545† | 1509† |
| Chess Puzzles | Epoch AI Benchmarking Hub | 67.0% | 51.0% |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 71.0% | 34.0% |
| CL-bench | Epoch AI Benchmarking Hub | – | 20.8%† |
| CL-bench Life | Epoch AI Benchmarking Hub | – | 16.9%† |
| LMCA | Epoch AI Benchmarking Hub | 68.2% | 53.8% |
| DTBench | Epoch AI Benchmarking Hub | 98.9% | 95.5% |
显示全部 221 项基准测试
| 指标 | 发布方 | Claude Opus 5.5 (max) | Gemini 3.1 Pro (high) |
|---|---|---|---|
| 推理 | |||
| ForecastBench | ForecastBench | – | 65.2% |
| 编程 | |||
| SciCode | SciCode | 66.9% | 58.7%† |
| WeirdML | WeirdML | – | 72.1%† |
| FrontierCode | Cognition | 54.6%† | – |
| GSO-Bench | GSO-Bench | – | 22.6%† |
| Arena Coding | Arena (LMArena) | 1548† | 1520† |
| Arena WebDev | Arena (LMArena) | 1815 | 1446† |
| LiveBench 2.0 Coding | LiveBench | 87.1% | – |
| LiveBench 1.0 Coding | LiveBench | 89.3% | 76.5% |
| LiveCodeBench | Vals AI | – | 88.5% |
| IOI | Vals AI | 95.1% | 51.8% |
| SWE-bench (Vals) | Vals AI | – | 78.8% |
| SWE-Bench Pro | Scale AI SEAL | – | 46.1%† |
| Code Migration | Vals AI | 66.7% | 17.3% |
| Code Migration v1.1 | Vals AI | 81.0% | – |
| ProgramBench | Vals AI | 18.5% | 0.0% |
| SWE-bench Multimodal | Vals AI | 70.2% | – |
| Vibe Code Bench 1-100 | Vals AI | 30.4%† | 6.7% |
| Vibe Code Bench v1.1 | Vals AI | 90.3% | 32.0% |
| Vibe Code Bench 1-100 v1.1 | Vals AI | 30.9% | – |
| SWE-Bench Pro (private) | Scale AI SEAL | – | 32.2%† |
| SWE Atlas v1.0: Codebase QnA | Scale AI SEAL | – | 13.5%† |
| SWE Atlas v1.0: Refactoring | Scale AI SEAL | – | 33.8%† |
| SWE Atlas v1.0: Test Writing | Scale AI SEAL | – | 29.8%† |
| SWE Atlas v1.1 | Scale AI SEAL | 77.4%† | – |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | 77.4%† | – |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | 77.8%† | – |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | 77.1%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 95.6% | 8.9% |
| DeepSWE v1.1 | Epoch AI Benchmarking Hub | – | 11.7%† |
| CursorBench | Cursor | 57.8% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 2147.4† | 1160.6† |
| AlgoTune | AlgoTune | – | 2.0† |
| FrontierSWE | Epoch AI Benchmarking Hub | 62.3% | – |
| WeirdML v3 | WeirdML | 31.2%† | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 59.6% | 4.0%† |
| Terminal-Bench 2.1 (AA) | Artificial Analysis | – | 73.8%† |
| Startrise Frontend | Startrise | 80.2%† | – |
| GameBench 2 Overall | DuelLab GameBench 2 | 87.6† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 94.8† | – |
| Design Arena Website | Design Arena | 1326† | 1226† |
| FrontierCode 1.1 (AA) | Artificial Analysis | 52.2% | – |
| FrontierCode 2.0 (AA) | Artificial Analysis | 46.4% | – |
| 智能体与工具 | |||
| Terminal-Bench | Terminal-Bench | 65.8% | 80.2%† |
| Remote Labor Index | Scale AI / CAIS | 21.3%† | – |
| APEX-Agents | Mercor | 73.5% | 35.1%† |
| Arena Agent | Arena (LMArena) | 19.5 | -4.5† |
| LiveBench Long Agentic | LiveBench | 72.8% | – |
| LiveBench Browser | LiveBench | 88.4% | – |
| LiveBench Agentic Coding | LiveBench | 71.7% | 44.1% |
| Terminal-Bench Hard | Artificial Analysis | – | 53.8%† |
| τ²-Bench Telecom (AA) | Artificial Analysis | – | 95.6%† |
| Terminal-Bench 2.1 (Vals) | Vals AI | 87.6%† | 70.8% |
| MCP Atlas | Scale AI SEAL | 85.4%† | 78.2% |
| HiL-Bench | Scale AI SEAL | 59.0%† | 35.3%† |
| τ³-bench Banking | τ²-bench | – | 26.0% |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 68.3% | 14.6%† |
| GDPval-AA v2.1 | Artificial Analysis | 1866 | 799† |
| τ³-Banking (AA) | Artificial Analysis | – | 21.4%† |
| ITBench SRE (AA) | Artificial Analysis | 38.2% | 30.3%† |
| Analyst Agent (AA) | Artificial Analysis | 56.3% | 41.3%† |
| APEX-Agents (AA) | Artificial Analysis | – | 32.0%† |
| CUA-bench | Vals AI | 14.0% | – |
| CyberBench | Vals AI | 55.4% | – |
| Finance Agent v2 | Vals AI | 58.6% | 43.0% |
| Finance Agent v2.1 | Vals AI | 67.8% | – |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 9.2% | – |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 3.8% | 0.0% |
| Legal Research Bench | Vals AI | 50.5% | 20.7% |
| Robo-bench | Vals AI | 14.6% | – |
| SREBench | Vals AI | 33.6% | 0.0% |
| Tax Agent Bench | Vals AI | 44.5% | 9.3% |
| Terminal-Bench 2.0 (Vals) | Vals AI | – | 67.4% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 65.2% | 2.5% |
| Terminal-Bench Science | Vals AI | 47.1% | 1.4% |
| Time Horizon Index: KSP | Vals AI | 91.3% | – |
| Web Search Index | Vals AI | 53.8% | – |
| Arena Search | Arena (LMArena) | 1213† | 1209† |
| Arena Agent Task Completion | Arena (LMArena) | 20.7 | -11.3† |
| BALROG | BALROG | – | 57.0%† |
| DeepResearch Bench | DeepResearch Bench | – | 47.8% |
| PostTrainBench | Epoch AI Benchmarking Hub | 43.8% | 18.1%† |
| ExploitBench | Epoch AI Benchmarking Hub | – | 26.1%† |
| METR Time Horizons | METR | – | 77.0%† |
| GBAEval | Epoch AI Benchmarking Hub | – | 0.8%† |
| Vending-Bench 2 | Andon Labs | 9235.3† | 911.2† |
| SpireClimb | Epoch AI Benchmarking Hub | 5.8% | – |
| AutomationBench | Artificial Analysis | 69.5% | 35.4%† |
| GDP.pdf (AA) | Artificial Analysis | 26.2% | 17.8%† |
| Harvey LAB-AA v1.1 | Harvey | 4.2% | 0.0%† |
| EnterpriseOps-Gym | Artificial Analysis | – | 42.2%† |
| AA-Briefcase v1.1 | Artificial Analysis | 1801 | 457† |
| GDPx | GDPx by SemiAnalysis | 12.0% | – |
| Design Arena Agentic | Design Arena | 1339† | 1088† |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 59.0% | – |
| CWE-Bench (AA) | Artificial Analysis | 58.3% | – |
| DeepSec-Bench (AA) | Artificial Analysis | 26.9% | – |
| CyberGym end-to-end (AA) | Artificial Analysis | 0.8% | – |
| MKA (AA) | Artificial Analysis | 53.0% | – |
| BrowseComp | OpenAI | – | 31.2%† |
| DeepSearchQA | – | 60.2%† | |
| FACTS Search | 88.4%† | 90.2%† | |
| 数学 | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 91.2% | 59.6%† |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 95.0% | 26.8%† |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 100.0% | 95.6% |
| ProofBench | Vals AI | 100.0% | 26.0%† |
| LiveBench Mathematics | LiveBench | 97.1% | 91.0% |
| AIME (Vals) | Vals AI | – | 98.1% |
| AIME 2026 | MathArena | – | 98.3% |
| HMMT February 2026 | MathArena | – | 94.7% |
| USAMO 2026 | MathArena | – | 74.4% |
| MathArena Apex | MathArena | – | 60.9% |
| ArXivMath | MathArena | 82.6%† | – |
| BrokenArXiv | MathArena | 77.3%† | – |
| Arena Maths | Arena (LMArena) | 1514† | 1488† |
| FrontierMath Erdős | Epoch AI Benchmarking Hub | 2.9% | – |
| 知识 | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 72.2% | 73.5% |
| LiveBench 2.0 Data Analysis | LiveBench | 80.3% | – |
| LiveBench Research | LiveBench | 88.9% | – |
| LiveBench 1.0 Data Analysis | LiveBench | 80.3% | 78.5% |
| AA-Omniscience v1.1 | Artificial Analysis | 48.1 | – |
| MMLU-Pro | Vals AI | – | 91.0% |
| LegalBench | Vals AI | – | 87.4% |
| CorpFin | Vals AI | – | 64.5% |
| TaxEval | Vals AI | – | 72.9% |
| MedQA | Vals AI | – | 96.4% |
| PRBench Finance | Scale AI SEAL | – | 41.9%† |
| PRBench Legal | Scale AI SEAL | 47.6%† | 44.0%† |
| MultiNRC | Scale AI SEAL | – | 64.7%† |
| BioMysteryBench | Vals AI | 79.3% | 60.0% |
| CaseLaw v2 | Vals AI | – | 64.8% |
| ClaimBench | Vals AI | 100.0% | – |
| Excel Modeling Benchmark | Vals AI | 75.9% | 52.6% |
| Excel Modeling Benchmark v1.1 | Vals AI | 75.8% | – |
| MedCode | Vals AI | 49.8% | 59.1% |
| MedScribe | Vals AI | 91.4% | 76.1% |
| Public Benefits Bench v1.1 | Vals AI | 70.6% | 53.8% |
| Access to Justice Benchmark | Vals AI | 73.8% | – |
| Financial Coaching Bench | Vals AI | 76.2% | 66.7% |
| Immigration Bench | Vals AI | 72.1% | – |
| DrugDiscoveryBench | Scale AI SEAL | – | 41.9% |
| EBR-bench | Epoch AI Benchmarking Hub | 71.4% | 14.3%† |
| GDP.pdf | Epoch AI Benchmarking Hub | 30.6% | 17.0% |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 65.5% | – |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 49.7% | – |
| Global-MMLU-Lite (AA) | Artificial Analysis | – | 93.2%† |
| HealthBench Professional (AA) | OpenAI | 59.9% | – |
| FACTS Parametric | 58.7%† | 79.4%† | |
| 指令遵循 | |||
| LiveBench Language | LiveBench | 86.3% | 85.4% |
| LiveBench Instruction Following | LiveBench | 65.7% | 79.1% |
| IFBench | Artificial Analysis | – | 77.1%† |
| MultiChallenge | Scale AI SEAL | – | 71.4%† |
| TutorBench | Scale AI SEAL | – | 53.0%† |
| Arena Instruction Following | Arena (LMArena) | 1518† | 1480† |
| 人类偏好 | |||
| Arena Text | Arena (LMArena) | 1507† | 1487† |
| EQ-Bench 4 | EQ-Bench | 1340† | 1132† |
| EQ-Bench Creative Writing v3 | EQ-Bench | 2051† | 1491† |
| Arena Creative Writing | Arena (LMArena) | 1517† | 1482† |
| Arena Multi-turn | Arena (LMArena) | 1502† | 1495† |
| Arena Longer Queries | Arena (LMArena) | 1520† | 1499† |
| 多模态 | |||
| Arena Vision | Arena (LMArena) | 1324† | 1296† |
| LiveBench Design | LiveBench | 83.3% | – |
| MMMU-Pro | Artificial Analysis | 87.7% | 82.4%† |
| MMMU-Pro (Vals) | Vals AI | – | 88.2% |
| MortgageTax | Vals AI | – | 69.4% |
| SAGE | Vals AI | 91.3% | 83.2% |
| Humanity's Sixth Sense | Scale AI SEAL | 44.6%† | – |
| VTB | Scale AI SEAL | – | 29.0%† |
| GeoBench | GeoBench | – | 84.0%† |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 51.2%† | 26.5%† |
| Furniture Assembly | Epoch AI Benchmarking Hub | 83.3% | 26.7% |
| VIALS-AA | Handshake AI | 61.5% | – |
| Code2Video-AA | Artificial Analysis | 1536 | – |
| AA-3D | Artificial Analysis | 1202 | – |
| FACTS Multimodal | 56.4%† | 50.4%† | |
| 长上下文 | |||
| AA-LCR v1.1 | Artificial Analysis | 84.7% | 82.0%† |
| Arena Document | Arena (LMArena) | – | 1459† |
| MLCR | Artificial Analysis | 27.8% | 11.1%† |
| FACTS Grounding | 62.5%† | 56.6%† | |
| 综合 | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 168.5† | 156.2† |
| LiveBench 2.0 | LiveBench | 83.4% | – |
| LiveBench 1.0 | LiveBench | 83.2% | 77.0% |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 59.1 | – |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 57.6 | 29.7† |
| AA Progress Index v1 | Artificial Analysis | 350 | 296.1† |
| Vals Index | Vals AI | 70.1 | – |
| Vals Finance Index | Vals AI | 65.3% | – |
| Vals RSI Index | Vals AI | 37.3% | – |
| Vals Science Index | Vals AI | 58.6% | 27.4% |
| Vals Multimodal Index | Vals AI | – | 56.1% |
| Vals Router Index | Vals AI | 64.2%† | – |
| Vals Legal Index | Vals AI | 72.4% | – |
| AA Cyber Index | Artificial Analysis | 28.7 | – |
| 安全与诚实 | |||
| FORTRESS | Scale AI SEAL | 5.8% | 29.8%† |
| MASK | Scale AI SEAL | – | 42.4%† |
| DistressBench | Scale AI SEAL | 75.6%† | 68.8%† |
数据截至 2026年11月16日。对比的是每个模型的最佳配置;每项分数的来源链接都在模型页面上。
DuelLab GameBench 2, release gb2-20261111-7b598bb13de4 (https://benchmarks.duellab.org/releases/gb2-20261111-7b598bb13de4/), CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Scores mapped to BenchLeader model names; not endorsed by DuelLab.
变化
- Claude Opus 5.5 (max):38 项新结果(LiveBench 2.0, LiveBench Long Agentic, LiveBench 2.0 Reasoning…)。
- Claude Opus 5.5 (max):1 项新结果(PostTrainBench);1 项分数修订。
- Gemini 3.1 Pro (high):4 项新结果(AIME 2026, HMMT February 2026, USAMO 2026…)。
- Claude Opus 5.5 (max):1 项新结果(MirrorCode);1 项分数修订。
- Claude Opus 5.5 (max):BenchLeader Index 69.3 → 68.8。
- Claude Opus 5.5 (max):2 项分数修订。
- Claude Opus 5.5 (max):1 项分数修订。
- Claude Opus 5.5 (max):1 项分数修订。
- Claude Opus 5.5 (max):2 项分数修订。
- Claude Opus 5.5 (max):2 项分数修订。
相关对比
- Claude Opus 5.5 vs Claude Fable 5.2
- Claude Opus 5.5 vs Gemini 4 Carbon
- Claude Opus 5.5 vs GPT-6 Astra
- Claude Opus 5.5 vs Claude Fable 5.1
- Gemini 3.1 Pro vs Gemini 4 Carbon
- Claude Opus 5.5 vs GPT-6.1 Sol
- Claude Opus 5.5 vs Claude Fable 5
- Claude Opus 5.5 vs GPT-5.6 Sol
排行榜: 综合最佳大模型 · 最适合编程的大模型 · 最适合智能体的大模型 · 便宜好用的大模型
Claude Opus 5.5 vs Gemini 3.1 Pro:常见问题
- Claude Opus 5.5 比 Gemini 3.1 Pro 更好吗?
- 质量上 Claude Opus 5.5 (max) 领先:68.4 比 60.4。BenchLeader Index 综合了所有第三方质量基准测试;截至 2026年11月16日,Claude Opus 5.5 (max) 总体领先,但请结合你的用途查看各类别得分。
- Claude Opus 5.5 和 Gemini 3.1 Pro 哪个更便宜?价格会变吗?
- 目前 Gemini 3.1 Pro 更便宜:每百万 token $4.50 比 $8.00(按每 1 个输出 token 对应 3 个输入 token 混合计算)。两者均未公布价格调整。
- 编程用 Claude Opus 5.5 还是 Gemini 3.1 Pro 更好?
- Claude Opus 5.5 在编程方面得分更高(类别得分 79 比 59,50 为平均水平)。在两者共同的最主要编程基准测试 IOI 上,Claude Opus 5.5 为 95.1%,Gemini 3.1 Pro 为 51.8%。
- Claude Opus 5.5 和 Gemini 3.1 Pro 哪个更快?
- Claude Opus 5.5 输出更快(输出 116 比 111 token/秒,Artificial Analysis 7 日中位数)。
- 智能体和工具调用用 Claude Opus 5.5 还是 Gemini 3.1 Pro 更好?
- Claude Opus 5.5 在智能体任务方面得分更高(类别得分 61 比 57,50 为平均水平)。
更多问题(4)
- 处理长文档哪个更好?
- 两者都约为 1M token(Claude Opus 5.5 为 1,000,000,Gemini 3.1 Pro 为 1,048,576)。
- 哪个延迟更低?
- Gemini 3.1 Pro 先开始回答:据 Artificial Analysis 测量,首个回答 token 的等待时间为 15.2 秒比 380.2 秒。这主要取决于推理强度:设置越高,回答前思考得越久。
- 第三方基准测试对 Claude Opus 5.5 和 Gemini 3.1 Pro 的看法一致吗?
- 在两者都有结果的 38 项基准测试中,Claude Opus 5.5 领先 34 项,Gemini 3.1 Pro 领先 4 项。所有同时评测两者的综合评分都认为 Claude Opus 5.5 更强。
- Claude Opus 5.5 是什么时候发布的?
- Claude Opus 5.5 于 2026年9月22日 发布。