Claude Haiku 5.5 vs Muse Wonder Full
完整资料:Claude Haiku 5.5(Anthropic)和 Muse Wonder Full(Meta),包括全部基准测试、价格和排名。
更新于 · 数据来自 11 个来源
截至 2026年11月16日,质量上 Muse Wonder Full 领先 Claude Haiku 5.5(BenchLeader Index 61.7 比 56.7)。在两者都有成绩的所有类别中,Muse Wonder Full 均领先。
要点(截至 2026年11月16日)
- Claude Haiku 5.5 的 BenchLeader Index 为 56.7,Muse Wonder Full 为 61.7(BenchLeader,2026年11月16日)。
- Claude Haiku 5.5 每百万 token 混合价格为 $0.200(标价,2026年11月16日)。
- 在两者都有结果的 21 项基准测试中,Muse Wonder Full 领先 20 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”。
主要差异
- 质量上 Muse Wonder Full (max) 领先:61.7 比 56.7。
- Muse Wonder Full (max) 在编程、知识、长上下文、推理方面更强。
概览
| 规格 | Claude Haiku 5.5 | Muse Wonder Full |
|---|---|---|
| 提供商 | Anthropic | Meta |
| 发布日期 | 2026年10月7日 | 2026年10月12日 |
| 上下文窗口(token) | 1,000,000 | 1,000,000 |
| 对比时的配置 | Claude Haiku 5.5 (xhigh) | Muse Wonder Full (max) |
| 开放权重 | 否 | 否 |
| 每百万 token 价格(输入 / 输出 / 缓存输入) | $0.100 / $0.500 / $0.010 | – |
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 Haiku 5.5 (xhigh) | Muse Wonder Full (max) | 领先者 |
|---|---|---|---|
| 推理 | 57.3 | 65.8 | Muse Wonder Full |
| 编程 | 58.7 | 69.9 | Muse Wonder Full |
| 智能体与工具 | – | 63.6 | 无法比较 |
| 数学 | 66.1 | – | 无法比较 |
| 知识 | 49.0 | 62.3 | Muse Wonder Full |
| 长上下文 | 63.8 | 66.6 | Muse Wonder Full |
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):智能体与工具, 数学, 指令遵循, 多模态, 人类偏好。
各来源的分歧
在两者都有结果的 21 项基准测试中,Claude Haiku 5.5 领先 1 项,Muse Wonder Full 领先 20 项。所有同时评测两者的综合评分都认为 Muse Wonder Full 更强。
| 综合 Index | Claude Haiku 5.5 (xhigh) | Muse Wonder Full (max) | 领先者 |
|---|---|---|---|
| BenchLeader Index | 56.7 | 61.7 | Muse Wonder Full |
| AA Intelligence Index v5.0.1 | 34.6 | 50.1 | Muse Wonder Full |
| Vals Index | 58.7† | 66.8 | Muse Wonder Full |
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 Haiku 5.5 (xhigh) | Muse Wonder Full (max) |
|---|---|---|
| 每百万 token 混合价格 | $0.200 | – |
| 每百万 token 输入 / 输出 | $0.100 / $0.500 | – / – |
| 每百万 token 缓存输入 | $0.010 | – |
| 每项任务价格:聊天回复 | $0.0002 | – |
| 每项任务价格:总结一份 30 页的报告 | $0.0015 | – |
| 每项任务价格:代码编辑 | $0.0014 | – |
| 每项任务价格:智能体编程会话 | $0.0080 | – |
| 每项任务价格:结构化信息提取 | $0.0003 | – |
| 智能体编程会话(输入大多命中缓存) | $0.0040 | – |
| 每项基准测试任务成本(AA Index) | $0.248 | – |
| 每项任务输出 token(AA Index) | 136k | 88k |
| 其中推理部分 | 97k | 55k |
| 每项任务耗时(AA Index) | 10 min | – |
| 输出速度 | 231 tok/s | – |
| 首次回答时间 | 50.5 秒 | – |
| 完整回答耗时(最近一次测量 2026年11月13日) | 110.1 秒 | – |
每个模型都按其排名最高的推理强度设置进行对比。Claude Haiku 5.5 (xhigh) 得分 56.7,首个回答 token 需 50 秒;Claude Haiku 5.5 (high) 为 55.3(18 秒),Claude Haiku 5.5 (max) 为 55.3(185 秒)。Muse Wonder Full (max) 得分 61.7。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 139 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 Haiku 5.5 (xhigh) | Muse Wonder Full (max) |
|---|---|---|---|
| BenchLeader Index | BenchLeader | 56.7 | 61.7 |
| 编程得分 | BenchLeader | 58.7 | 69.9 |
| 知识得分 | BenchLeader | 49.0 | 62.3 |
| 长上下文得分 | BenchLeader | 63.8 | 66.6 |
| 数学得分 | BenchLeader | 66.1 | – |
| 推理得分 | BenchLeader | 57.3 | 65.8 |
| 智能体与工具得分 | BenchLeader | – | 63.6 |
| 混合价格(美元/百万 token) | BenchLeader | $0.200 | – |
| 输出速度 | BenchLeader | 231 tok/s | – |
| 首次回答时间 | BenchLeader | 50.5 秒 | – |
| 上下文窗口 | BenchLeader | 1,000,000 token | 1,000,000 token |
| 推理 | |||
| GPQA Diamond | Epoch AI Benchmarking Hub | 89.7%† | – |
| Humanity's Last Exam | Scale AI / CAIS | 30.1%† | – |
| Arena Hard Prompts | Arena (LMArena) | 1449† | – |
| LiveBench 2.0 Reasoning | LiveBench | 82.4%† | – |
| LiveBench 1.0 Reasoning | LiveBench | 81.2% | – |
| Humanity's Last Exam (AA) | Artificial Analysis | 42.7% | 53.6% |
| HLE-Diamond (AA) | Artificial Analysis | 28.8% | 48.5% |
| ARC-AGI-1 | ARC Prize | 87.5% | – |
| ARC-AGI-2 | ARC Prize | 61.3% | – |
| ARC-AGI-3 | ARC Prize | 2.8%† | – |
| CritPt | Artificial Analysis | 22.6% | 32.9% |
| CritPt v2 hard subset (AA) | Artificial Analysis | 47.5% | 56.5% |
| MysteryMechanism | Vals AI | 35.1%† | 35.1% |
| Arena Expert | Arena (LMArena) | 1465† | – |
| Chess Puzzles | Epoch AI Benchmarking Hub | 55.0%† | – |
| Mystery Game Puzzles | Epoch AI Benchmarking Hub | 30.0%† | – |
| LMCA | Epoch AI Benchmarking Hub | 48.0% | – |
| DTBench | Epoch AI Benchmarking Hub | 87.2% | – |
| 编程 | |||
| SciCode | SciCode | 51.7% | – |
| FrontierCode | Cognition | 46.4%† | – |
| Arena Coding | Arena (LMArena) | 1489† | – |
| Arena WebDev | Arena (LMArena) | 1593† | 1724 |
| LiveBench 2.0 Coding | LiveBench | 77.6%† | – |
| LiveBench 1.0 Coding | LiveBench | 76.4% | – |
| SciCode (AA) | Artificial Analysis | 51.7% | 62.6% |
显示全部 139 项基准测试
| 指标 | 发布方 | Claude Haiku 5.5 (xhigh) | Muse Wonder Full (max) |
|---|---|---|---|
| 编程 | |||
| IOI | Vals AI | 47.3%† | 79.8% |
| Code Migration | Vals AI | 44.6%† | 61.0% |
| Code Migration v1.1 | Vals AI | 52.4%† | 62.8% |
| ProgramBench | Vals AI | 2.5%† | 0.5% |
| SWE-bench Multimodal | Vals AI | 44.0%† | 47.3% |
| Vibe Code Bench 1-100 | Vals AI | 20.4%† | 34.0% |
| Vibe Code Bench v1.1 | Vals AI | 90.4%† | 86.0% |
| SWE Atlas v1.1 | Scale AI SEAL | 59.4%† | – |
| SWE Atlas v1.1: Codebase QnA | Scale AI SEAL | 62.1%† | – |
| SWE Atlas v1.1: Test Writing | Scale AI SEAL | 63.3%† | – |
| SWE Atlas v1.1: Refactoring | Scale AI SEAL | 52.9%† | – |
| SWE-Bench Pro V2 | Scale AI SEAL | 87.2%† | – |
| SWE-Bench Pro V2 HARD | Scale AI SEAL | 52.9%† | – |
| MirrorCode | Epoch AI Benchmarking Hub | 78.9%† | – |
| CursorBench | Cursor | 44.3% | – |
| ALE-Bench | Epoch AI Benchmarking Hub | 1587† | – |
| FrontierSWE | Epoch AI Benchmarking Hub | 43.8%† | – |
| WeirdML v3 | WeirdML | 11.7% | – |
| Terminal-Bench 4.0 (AA) | Artificial Analysis | 29.3% | 57.1% |
| GameBench 2 Overall | DuelLab GameBench 2 | 56.0† | – |
| GameBench 2 by reasoning setting | DuelLab GameBench 2 | 75.1 | – |
| Design Arena Website | Design Arena | 1253† | – |
| FrontierCode 1.1 (AA) | Artificial Analysis | 45.5% | 44.9% |
| FrontierCode 2.0 (AA) | Artificial Analysis | 29.9% | 37.8% |
| 智能体与工具 | |||
| Terminal-Bench | Terminal-Bench | 40.0%† | – |
| APEX-Agents | Mercor | 56.1%† | – |
| Arena Agent | Arena (LMArena) | 0.5† | 9 |
| LiveBench Long Agentic | LiveBench | 47.9%† | – |
| LiveBench Browser | LiveBench | 69.5%† | – |
| LiveBench Agentic Coding | LiveBench | 51.4% | – |
| GDPval-AA v2.1, normalised score | Artificial Analysis | 50.7% | 67.8% |
| GDPval-AA v2.1 | Artificial Analysis | 1523 | 1857 |
| Analyst Agent (AA) | Artificial Analysis | 53.8%† | – |
| CyberBench | Vals AI | 75.5%† | 73.6% |
| Finance Agent v2 | Vals AI | 54.1%† | 59.1% |
| Finance Agent v2.1 | Vals AI | 65.4%† | 67.6% |
| Harvey's Legal Agent Benchmark v1.1 | Vals AI | 4.2%† | 23.3% |
| Harvey's Legal Agent Benchmark v1 | Vals AI | 1.3%† | 21.7% |
| Legal Research Bench | Vals AI | 43.8%† | 53.4% |
| Robo-bench | Vals AI | – | 10.4% |
| SREBench | Vals AI | 5.7%† | 11.8% |
| Tax Agent Bench | Vals AI | 31.9%† | 43.0% |
| Terminal-Bench 4.0 (Vals) | Vals AI | 36.4%† | 58.1% |
| Terminal-Bench Science | Vals AI | 10.0%† | 37.1% |
| Arena Agent Task Completion | Arena (LMArena) | 1.6† | 11.1 |
| Vending-Bench 2 | Andon Labs | 3602.5† | – |
| AutomationBench | Artificial Analysis | 36.0% | 71.3% |
| GDP.pdf (AA) | Artificial Analysis | 18.2% | 32.6% |
| Harvey LAB-AA v1.1 | Harvey | 2.8% | 13.1% |
| AA-Briefcase v1.1 | Artificial Analysis | 1523 | 1729 |
| GDPx | GDPx by SemiAnalysis | 4.9%† | 9.0% |
| Terminal-Bench-Science 0.1 (AA) | Artificial Analysis | 18.1% | 37.1% |
| MKA (AA) | Artificial Analysis | 41.0%† | – |
| FACTS Search | 80.3%† | – | |
| 数学 | |||
| FrontierMath Tiers 1–3 | Epoch AI Benchmarking Hub | 75.1%† | – |
| FrontierMath Tier 4 | Epoch AI Benchmarking Hub | 46.3%† | – |
| OTIS Mock AIME | Epoch AI Benchmarking Hub | 98.9% | – |
| ProofBench | Vals AI | 89.0%† | – |
| LiveBench Mathematics | LiveBench | 92.8% | – |
| Arena Maths | Arena (LMArena) | 1451† | – |
| 知识 | |||
| SimpleQA Verified | Epoch AI Benchmarking Hub | 23.8%† | – |
| LiveBench 2.0 Data Analysis | LiveBench | 53.1%† | – |
| LiveBench Research | LiveBench | 84.3%† | – |
| LiveBench 1.0 Data Analysis | LiveBench | 72.8% | – |
| AA-Omniscience v1.1 | Artificial Analysis | 7.8 | 30.6 |
| LegalBench | Vals AI | – | 87.4% |
| BioMysteryBench | Vals AI | 70.4%† | 73.3% |
| ClaimBench | Vals AI | 100.0%† | – |
| Excel Modeling Benchmark | Vals AI | 59.0%† | 71.9% |
| Excel Modeling Benchmark v1.1 | Vals AI | 65.5%† | 75.0% |
| MedCode | Vals AI | 51.5%† | 49.9% |
| MedScribe | Vals AI | 89.5%† | 89.3% |
| Public Benefits Bench v1.1 | Vals AI | 63.2%† | 73.8% |
| Access to Justice Benchmark | Vals AI | – | 74.7% |
| Financial Coaching Bench | Vals AI | – | 75.0% |
| Immigration Bench | Vals AI | 61.8%† | 71.2% |
| EBR-bench | Epoch AI Benchmarking Hub | 43.8%† | – |
| AA-Omniscience v1.1: accuracy | Artificial Analysis | 33.8% | 41.0% |
| AA-Omniscience v1.1: non-hallucination | Artificial Analysis | 60.9% | 82.4% |
| HealthBench Professional (AA) | OpenAI | 58.4%† | – |
| FACTS Parametric | 13.1%† | – | |
| 指令遵循 | |||
| LiveBench Language | LiveBench | 63.5% | – |
| LiveBench Instruction Following | LiveBench | 66.5% | – |
| MultiChallenge | Scale AI SEAL | 67.7%† | – |
| Arena Instruction Following | Arena (LMArena) | 1426† | – |
| 人类偏好 | |||
| Arena Text | Arena (LMArena) | 1421† | – |
| Arena Creative Writing | Arena (LMArena) | 1361† | – |
| Arena Multi-turn | Arena (LMArena) | 1412† | – |
| Arena Longer Queries | Arena (LMArena) | 1419† | – |
| 多模态 | |||
| LiveBench Design | LiveBench | 79.3%† | – |
| SAGE | Vals AI | 89.8%† | – |
| Blueprint-Bench 2 | Epoch AI Benchmarking Hub | 33.0%† | – |
| Furniture Assembly | Epoch AI Benchmarking Hub | 47.5%† | – |
| VIALS-AA | Handshake AI | 49.1%† | – |
| Code2Video-AA | Artificial Analysis | 1192† | – |
| AA-3D | Artificial Analysis | 998† | – |
| FACTS Multimodal | 38.9%† | – | |
| 长上下文 | |||
| AA-LCR v1.1 | Artificial Analysis | 78.3% | 83.7% |
| MLCR | Artificial Analysis | 15.0% | 22.2% |
| FACTS Grounding | 57.0%† | – | |
| 综合 | |||
| Epoch Capabilities Index | Epoch AI Benchmarking Hub | 160.4† | – |
| LiveBench 2.0 | LiveBench | 68.9%† | – |
| LiveBench 1.0 | LiveBench | 72.1% | – |
| AA Intelligence Index v5.0.1 | Artificial Analysis | 34.6 | 50.1 |
| AA Intelligence Index v4.3.2 | Artificial Analysis | 41.3 | 55.8 |
| AA Progress Index v1 | Artificial Analysis | 322.2 | 343.2 |
| Vals Index | Vals AI | 58.7† | 66.8 |
| Vals Finance Index | Vals AI | 57.5%† | 64.5% |
| Vals RSI Index | Vals AI | – | 24.4% |
| Vals Science Index | Vals AI | 38.5%† | 48.5% |
| Vals Router Index | Vals AI | 58.7%† | 66.8%† |
| Vals Legal Index | Vals AI | 70.3%† | 77.9% |
| 安全与诚实 | |||
| FORTRESS | Scale AI SEAL | 9.5%† | – |
| DistressBench | Scale AI SEAL | 71.7%† | – |
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
- Muse Wonder Full (max):BenchLeader Index 62.6 → 61.7。
- Claude Haiku 5.5 (xhigh):1 项新结果(GameBench 2 by reasoning setting)。
- Claude Haiku 5.5 (xhigh):BenchLeader Index 57.3 → 56.8。
- Claude Haiku 5.5 (xhigh):BenchLeader Index 56.8 → 57.3。
- Claude Haiku 5.5 (xhigh):1 项分数修订。
- Muse Wonder Full (max):1 项分数修订。
- Muse Wonder Full (max):1 项分数修订。
- Muse Wonder Full (max):1 项新结果(Arena Agent)。
- Claude Haiku 5.5 (xhigh):1 项新结果(AA-Briefcase v1.1)。
- Claude Haiku 5.5 (xhigh):BenchLeader Index 58.4 → 56.8。
相关对比
- Claude Haiku 5.5 vs Claude Opus 5.5
- Claude Haiku 5.5 vs GPT-6 Astra
- Claude Haiku 5.5 vs Claude Fable 5.1
- Muse Wonder Full vs Claude Fable 5.2
- Muse Wonder Full vs Gemini 4 Carbon
- Muse Wonder Full vs Claude Opus 5.5
- Muse Wonder Full vs GPT-6 Astra
- Claude Sonnet 5.5 vs Muse Wonder Full
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 Haiku 5.5 vs Muse Wonder Full:常见问题
- Muse Wonder Full 比 Claude Haiku 5.5 更好吗?
- 质量上 Muse Wonder Full (max) 领先:61.7 比 56.7。BenchLeader Index 综合了所有第三方质量基准测试;截至 2026年11月16日,Muse Wonder Full (max) 总体领先,但请结合你的用途查看各类别得分。
- 编程用 Claude Haiku 5.5 还是 Muse Wonder Full 更好?
- Muse Wonder Full 在编程方面得分更高(类别得分 70 比 59,50 为平均水平)。在两者共同的最主要编程基准测试 FrontierCode 2.0 (AA) 上,Claude Haiku 5.5 为 29.9%,Muse Wonder Full 为 37.8%。
- 处理长文档哪个更好?
- Muse Wonder Full 在长上下文基准测试上得分更高(67 比 64)。两者都支持 1,000,000 token 的上下文。
- 第三方基准测试对 Claude Haiku 5.5 和 Muse Wonder Full 的看法一致吗?
- 在两者都有结果的 21 项基准测试中,Claude Haiku 5.5 领先 1 项,Muse Wonder Full 领先 20 项。所有同时评测两者的综合评分都认为 Muse Wonder Full 更强。
- Muse Wonder Full 是什么时候发布的?
- Muse Wonder Full 于 2026年10月12日 发布。