Claude Fable 5.1vsGPT-5.6 Sol
在 41 个同模式 benchmark 中，Claude Fable 5.1 整体领先：Claude Fable 5.1 领先 34 项，GPT-5.6 Sol 领先 7 项，持平 0 项，平均分差 +12.50。另有 8 项测试模式不同，仅供参考。
Claude Fable 5.1
Anthropic · 2026-09-01 · 推理大模型
GPT-5.6 Sol
OpenAI · 2026-06-26 · 推理大模型
Claude Fable 5.134 项(83%)(17%)7 项GPT-5.6 Sol
评测分数
按能力类目分组，每组内按分差大小排列；共 41 项。
仓库修复与多文件工程
Claude Fable 5.1 领先 2/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| FrontierSWE v2 | 56.294 / 22Max (With Tools) | 32.209 / 22Max (With Tools) | +24.09 |
| APEX-SWE | 63.604 / 73Max (With Tools) | 4818 / 73Max (With Tools) | +15.60 |
| DeepSWE | 67.4035 / 100Max (With Tools) | 72.6716 / 100Max (With Tools) | -5.27 |
客服与业务流程
GPT-5.6 Sol 领先 2/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Public Benefits Bench v1.1 | 74.902 / 26Max (With Tools) | 66.5110 / 26Max (With Tools) | +8.39 |
| SAGE | 48.5320 / 65Max (With Tools) | 52.564 / 65Max (With Tools) | -4.03 |
| τ³-Banking | 45.8014 / 168Extra-High (With Tools) | 46.9113 / 168Extra-High (With Tools) | -1.11 |
数学
Claude Fable 5.1 领先 3/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| ProofBench v1.1 (Lean 4) | 1001 / 29Max (With Tools) | 837 / 29Max (With Tools) | +17 |
| FrontierMath Tier 4 v2 | 87.8010 / 71Max (With Tools) | 82.9313 / 71Max (With Tools) | +4.88 |
| FrontierMath v2 | 90.184 / 114Max (With Tools) | 89.126 / 114Max (With Tools) | +1.05 |
法律
Claude Fable 5.1 领先 3/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Legal Research Bench | 55.291 / 46Max (With Tools) | 48.088 / 46Max (With Tools) | +7.21 |
| Harvey Lab-AA | 93.026 / 44Max (With Tools) | 87.1817 / 44Max (With Tools) | +5.84 |
| Harvey's Legal Agent Benchmark | 6.6720 / 50Max (With Tools) | 2.5035 / 50Max (With Tools) | +4.17 |
自主开发与终端任务
Claude Fable 5.1 领先 3/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Terminal-Bench 4.0 | 57.886 / 104Max (With Tools) | 37.2727 / 104Max (With Tools) | +20.61 |
| CursorBench 4.0 | 51.801 / 43Max (With Tools) | 41.7010 / 43Max (With Tools) | +10.10 |
| CursorBench 3.2 | 73.401 / 5Max (With Tools) | 67.204 / 5Max (With Tools) | +6.20 |
金融
Claude Fable 5.1 领先 3/3| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Tax Agent Bench | 77.641 / 21Max (With Tools) | 67.958 / 21Max (With Tools) | +9.69 |
| Finance Agent v2 | 58.888 / 46Max (With Tools) | 53.7624 / 46Max (With Tools) | +5.12 |
| EMB (Excel Modeling) | 76.671 / 46Max (With Tools) | 72.347 / 46Max (With Tools) | +4.33 |
临床工作与医疗决策
Claude Fable 5.1 领先 2/2| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| MedCode | 53.516 / 66Max (With Tools) | 43.9733 / 66Max (With Tools) | +9.54 |
| MedScribe | 91.292 / 68Max (With Tools) | 85.2320 / 68Max (With Tools) | +6.06 |
文档与图表理解
胶着 2/2| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Chartography | 86.203 / 16Max (With Tools) | 79.906 / 16Max (With Tools) | +6.30 |
| GDP.pdf | 26.8013 / 124Thinking High (No Tools) | 27.8010 / 124Thinking High (No Tools) | -1 |
科学知识与推理
胶着 2/2| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| MysteryMechanism | 47.753 / 14Max (With Tools) | 33.337 / 14Max (With Tools) | +14.42 |
| CritPt | 29.7011 / 204Max (No Tools) | 32.301 / 204Max (No Tools) | -2.60 |
科学计算与科研编程
Claude Fable 5.1 领先 2/2| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Terminal-Bench-Science 0.1 | 404 / 18Max (With Tools) | 22.406 / 18Max (With Tools) | +17.60 |
| SciCode | 63.102 / 135Max (No Tools) | 57.1022 / 135Max (No Tools) | +6 |
程序生成与编辑
Claude Fable 5.1 领先 2/2| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Program Bench | 87.601 / 17Max (With Tools) | 2310 / 17Max (With Tools) | +64.60 |
| Vibe Code Bench v1.1 | 90.265 / 66Max (With Tools) | 80.5019 / 66Max (With Tools) | +9.76 |
专业数据分析
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| AA-AnalystAgent | 57.502 / 29Max (With Tools + Internet) | 47.507 / 29Max (With Tools + Internet) | +10 |
写作与创意表达
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Creative Writing | 2,1622 / 112Normal (No Tools) | 1,9729 / 112Normal (No Tools) | +190.40 |
办公与文档流程
GPT-5.6 Sol 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| AutomationBench | 31.4021 / 28Max (With Tools) | 45.8011 / 28Max (With Tools) | -14.40 |
多轮记忆与持续上下文
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| EBR-bench | 57.143 / 23Max (With Tools) | 44.766 / 23Max (With Tools) | +12.38 |
抽象归纳与泛化
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| ARC-AGI-1 | 97.5011 / 226Max (No Tools) | 96.5024 / 226Max (No Tools) | +1 |
机器学习工程
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| WeirdML v3 | 0.264 / 21Extra-High (With Tools) | 0.159 / 21Extra-High (With Tools) | +0.11 |
桌面工作流
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| OSWorld 2.0 | 77.901 / 15Max (With Tools) | 62.609 / 15Max (With Tools) | +15.30 |
算法与竞赛编程
GPT-5.6 Sol 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| IOI (Vals v2) | 90.786 / 30Max (With Tools) | 91.175 / 30Max (With Tools) | -0.39 |
维护、优化与工程管理
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Code Migration | 54.619 / 48Max (With Tools) | 52.9210 / 48Max (With Tools) | +1.69 |
综合知识考试
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| HLE | 60.9010 / 239Max (No Tools) | 44.5068 / 239Max (No Tools) | +16.40 |
跨应用与工具编排
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| MCP-Atlas | 87.204 / 51Max (With Tools) | 81.8014 / 51Max (With Tools) | +5.40 |
跨能力综合测试
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| LiveBench | 83.411 / 157Max (No Tools) | 81.0510 / 157Max (No Tools) | +2.36 |
跨能力聚合指数
Claude Fable 5.1 领先 1/1| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol | 分差 |
|---|---|---|---|
| Vals Index v2.1 | 65.834 / 37Max (task-specific tools) | 58.019 / 37Max (task-specific tools) | +7.82 |
测试模式不同的成绩
共 8 项。两款模型的公开成绩来自不同测试模式，分数不直接可比，不计入胜负和平均分差。
| 评测项 | Claude Fable 5.1 | GPT-5.6 Sol |
|---|---|---|
| Vals Index | 68.83Max (With Tools) | 72.63Extra-High (No Tools) |
| FrontierCode 1.1 Extended | 63.60Thinking Medium (With Tools) | 60.60Max (With Tools) |
| FrontierCode 1.1 Main | 50.90Thinking Medium (With Tools) | 47.50Reported best (effort unspecified) |
| SimpleBench | 86.60Thinking (No Tools) | 64.80Extra-High (No Tools) |
| HealthBench Professional (length-adjusted) | 62.10Max (No Tools) | 60.50Reported best (effort unspecified) |
| SWE-Bench Pro - Public | 81.20Max (With Tools) | 64.60Extra-High (With Tools) |
| SWE-Bench Pro V2 | 99.10Thinking High (With Tools) | 95.50Extra-High (With Tools) |
| SWE-Bench Pro V2 Hard | 92.20Thinking High (With Tools) | 82.40Extra-High (With Tools) |
规格对比
| 字段 | Claude Fable 5.1 | GPT-5.6 Sol |
|---|---|---|
| 发布机构 | Anthropic | OpenAI |
| 发布时间 | 2026-09-01 | 2026-06-26 |
| 模型类型 | 推理大模型 | 推理大模型 |
| 架构 | 稠密模型 | 稠密模型 |
| 参数规模 | 暂无数据 | 暂无数据 |
| 上下文长度 | 1M | 1.05M |
| 最大输出 | 128K | 128K |
API 调用价格
价格优先使用 DataLearner 配置的 API 记录；缺失项不做推测。
| 价格项 | Claude Fable 5.1 | GPT-5.6 Sol |
|---|---|---|
| 文本输入 | $10 / 1M tokens | $4 / 1M tokens |
| 文本输出 | $50 / 1M tokens | $20 / 1M tokens |
| 缓存读取 | $0.25 / 1M tokens | $0.4 / 1M tokens |
| 缓存写入 | $12.5 / 1M tokens | $5 / 1M tokens |
小结
- Claude Fable 5.1在以下类目领先:仓库修复与多文件工程 (2/3)、数学 (3/3)、法律 (3/3)、自主开发与终端任务 (3/3)、金融 (3/3)、临床工作与医疗决策 (2/2)、科学计算与科研编程 (2/2)、程序生成与编辑 (2/2)、专业数据分析 (1/1)、写作与创意表达 (1/1)、多轮记忆与持续上下文 (1/1)、抽象归纳与泛化 (1/1)、机器学习工程 (1/1)、桌面工作流 (1/1)、维护、优化与工程管理 (1/1)、综合知识考试 (1/1)、跨应用与工具编排 (1/1)、跨能力综合测试 (1/1)、跨能力聚合指数 (1/1)
- GPT-5.6 Sol在以下类目领先:客服与业务流程 (2/3)、办公与文档流程 (1/1)、算法与竞赛编程 (1/1)
- 胶着类目:文档与图表理解、科学知识与推理
41 个同模式 benchmark 上，Claude Fable 5.1 平均高出 12.50 分。
单项差距最大的 benchmark：Creative Writing — Claude Fable 5.1 2,162，GPT-5.6 Sol 1,972（分差 +190.40）。
本页正文由结构化模型、价格与 benchmark 数据生成，不使用实时 LLM 撰写。