EN中
独立模型评测对比

哪款 AI 模型表现领先？

真实评测成绩，一眼看懂。按任务选榜单、看模型差异、读懂评分。

15 组图表3 个开放数据来源 · 日期截至 2026-10-23
综合↗
Gemini 4 Argon (High)
1525.7Arena Rating · Arena
编程↗
Claude Opus 5.5 (High)
1545Arena Rating · Arena
推理↗
Claude Opus 5.5 (High)
1538.5Arena Rating · Arena
图像↗
gpt-image-2.5-sunburst
1424.6Arena Rating · Arena
按需求查找

探索全部评测图表

根据编程、推理、图像等任务选图。每张卡片都能进入交互对比。

15 / 15 组图表
Arena2026-10-22

综合模型偏好

#1Gemini 4 Argon (High)1525.7
General model preference. Zoomed horizontal axis, from 1498.7 to 1530.3Gemini 4 Argon (High) · 1525.7 Arena Rating · zoomed horizontal axisClaude Opus 5.5 (High) · 1516.2 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1512.5 Arena Rating · zoomed horizontal axisClaude Opus 5 (Max) · 1508.3 Arena Rating · zoomed horizontal axisClaude Opus 5 (High) · 1503.3 Arena Rating · zoomed horizontal axis1498.71530.3ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-22

数学推理

#1Gemini 4 Argon (High)1537.8
Mathematical reasoning. Zoomed horizontal axis, from 1526.2 to 1542.4Gemini 4 Argon (High) · 1537.8 Arena Rating · zoomed horizontal axisClaude Opus 5 (Max) · 1533 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1532.3 Arena Rating · zoomed horizontal axisGemini 3.8 Flash (High) · 1531.6 Arena Rating · zoomed horizontal axisClaude Opus 5 (High) · 1530.8 Arena Rating · zoomed horizontal axis1526.21542.4ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-22

高难度问题

#1Claude Opus 5.5 (High)1538.5
Hard prompts. Zoomed horizontal axis, from 1518.7 to 1543.1Claude Opus 5.5 (High) · 1538.5 Arena Rating · zoomed horizontal axisGemini 4 Argon (High) · 1534.4 Arena Rating · zoomed horizontal axisClaude Opus 4.6 (High) · 1529.9 Arena Rating · zoomed horizontal axisClaude Opus 4.6 · 1527.5 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1523.3 Arena Rating · zoomed horizontal axis1518.71543.1ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-22

编程偏好

#1Claude Opus 5.5 (High)1545
Coding preference. Zoomed horizontal axis, from 1527.9 to 1549.6Claude Opus 5.5 (High) · 1545 Arena Rating · zoomed horizontal axisGemini 4 Argon (High) · 1539.3 Arena Rating · zoomed horizontal axisClaude Opus 4.6 · 1534.4 Arena Rating · zoomed horizontal axisClaude Opus 5 (High) · 1532.6 Arena Rating · zoomed horizontal axisMiMo-V2.6-Pro · 1532.5 Arena Rating · zoomed horizontal axis1527.91549.6ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-22

网页开发

#1Claude Opus 5.5 (Max)1806.9
Web development. Zoomed horizontal axis, from 1733.9 to 1816.4Claude Opus 5.5 (Max) · 1806.9 Arena Rating · zoomed horizontal axisGPT-6 Astra (Max) · 1774.7 Arena Rating · zoomed horizontal axisClaude Sonnet 5.5 (xHigh) · 1767.5 Arena Rating · zoomed horizontal axisGPT-6.1 Sol (Max) · 1749 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1743.4 Arena Rating · zoomed horizontal axis1733.91816.4ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-22

中文对话

#1Gemini 4 Argon (High)1600.2
Chinese conversations. Zoomed horizontal axis, from 1562.5 to 1605.1Gemini 4 Argon (High) · 1600.2 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1583.6 Arena Rating · zoomed horizontal axisClaude Opus 5.5 (High) · 1581 Arena Rating · zoomed horizontal axisClaude Opus 5 (Max) · 1573.1 Arena Rating · zoomed horizontal axisClaude Opus 5 (High) · 1567.4 Arena Rating · zoomed horizontal axis1562.51605.1ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-19

视觉理解

#1Claude Fable 5 (High)1324.9
Image understanding. Zoomed horizontal axis, from 1312.9 to 1328.9Claude Fable 5 (High) · 1324.9 Arena Rating · zoomed horizontal axisClaude Opus 5.5 (High) · 1320.8 Arena Rating · zoomed horizontal axisClaude Opus 5 (High) · 1319.6 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1319.5 Arena Rating · zoomed horizontal axisClaude Opus 4.7 · 1316.9 Arena Rating · zoomed horizontal axis1312.91328.9ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-07

文本生成图像

#1gpt-image-2.5-sunburst1424.6
Text-to-image. Zoomed horizontal axis, from 1318.2 to 1438.5gpt-image-2.5-sunburst · 1424.6 Arena Rating · zoomed horizontal axisgpt-image-2.5-flare · 1397.6 Arena Rating · zoomed horizontal axisgpt-image-2 (medium) · 1383 Arena Rating · zoomed horizontal axisgrok-imagine-image-2.0 (canvas) · 1333.5 Arena Rating · zoomed horizontal axismai-image-2.6 · 1332.1 Arena Rating · zoomed horizontal axis1318.21438.5ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-20

图像编辑

#1GPT-Image-2.5 Sunburst1514.4
Image editing. Zoomed horizontal axis, from 1413.0 to 1527.6GPT-Image-2.5 Sunburst · 1514.4 Arena Rating · zoomed horizontal axisGPT-Image-2.5 Flare · 1480.7 Arena Rating · zoomed horizontal axisGPT-Image-2 (Medium) · 1460.6 Arena Rating · zoomed horizontal axisGrok Imagine Image 2.0 · 1439.4 Arena Rating · zoomed horizontal axisGemini Nano Banana 2.1 · 1426.2 Arena Rating · zoomed horizontal axis1413.01527.6ZOOMED · NOT ZERO
Arena Rating查看图表
Arena2026-10-20

智能体能力 · IPS

#1Claude Opus 5.5 (Max)0.2092
Agent capability · IPS. Zoomed horizontal axis, from 0.1143 to 0.2216Claude Opus 5.5 (Max) · 0.2092 IPS · zoomed horizontal axisGPT-6 Astra (Max) · 0.147 IPS · zoomed horizontal axisClaude Fable 5.1 (Max) · 0.1417 IPS · zoomed horizontal axisGPT-6.1 Sol (Max) · 0.1355 IPS · zoomed horizontal axisClaude Sonnet 5.5 (Max) · 0.1267 IPS · zoomed horizontal axis0.11430.2216ZOOMED · NOT ZERO
IPS查看图表
Arena2026-10-22

指令遵循

#1Claude Opus 5.5 (High)1538.8
Instruction following. Zoomed horizontal axis, from 1513.8 to 1543.4Claude Opus 5.5 (High) · 1538.8 Arena Rating · zoomed horizontal axisGemini 4 Argon (High) · 1527.6 Arena Rating · zoomed horizontal axisClaude Opus 4.6 (High) · 1521 Arena Rating · zoomed horizontal axisClaude Fable 5.1 (Max) · 1518.9 Arena Rating · zoomed horizontal axisClaude Opus 5 (Max) · 1518.4 Arena Rating · zoomed horizontal axis1513.81543.4ZOOMED · NOT ZERO
Arena Rating查看图表
MTEB2026-10-23

科学文献检索 · SciFact

#1Qwen/Qwen3-Embedding-4B78.33
Scientific retrieval · SciFact. Zoomed horizontal axis, from 73.26 to 78.99Qwen/Qwen3-Embedding-4B · 78.33 nDCG@10 × 100 · zoomed horizontal axisAlibaba-NLP/gte-modernbert-base · 76.41 nDCG@10 × 100 · zoomed horizontal axisBAAI/bge-large-en-v1.5 · 74.64 nDCG@10 × 100 · zoomed horizontal axisBAAI/bge-base-en-v1.5 · 74.34 nDCG@10 × 100 · zoomed horizontal axismixedbread-ai/mxbai-embed-large-v1 · 73.92 nDCG@10 × 100 · zoomed horizontal axis73.2678.99ZOOMED · NOT ZERO
nDCG@10 × 100查看图表
MTEB2026-10-23

语义相似度 · STS

#1Qwen/Qwen3-Embedding-4B93.7
Semantic similarity · STS. Zoomed horizontal axis, from 86.59 to 94.63Qwen/Qwen3-Embedding-4B · 93.7 Spearman × 100 · zoomed horizontal axisQwen/Qwen3-Embedding-0.6B · 91.13 Spearman × 100 · zoomed horizontal axisjinaai/jina-embeddings-v3 · 89.43 Spearman × 100 · zoomed horizontal axismixedbread-ai/mxbai-embed-large-v1 · 89.29 Spearman × 100 · zoomed horizontal axisBAAI/bge-large-en-v1.5 · 87.52 Spearman × 100 · zoomed horizontal axis86.5994.63ZOOMED · NOT ZERO
Spearman × 100查看图表
EvalPlus2026-10-23

HumanEval+ 代码测试

#1O1 Mini (Sept 2024)89
HumanEval+ code tests. Zoomed horizontal axis, from 86.24 to 89.36O1 Mini (Sept 2024) · 89 pass@1 (%) · zoomed horizontal axisO1 Preview (Sept 2024) · 89 pass@1 (%) · zoomed horizontal axisGPT 4o (Aug 2024) · 87.2 pass@1 (%) · zoomed horizontal axisQwen2.5-Coder-32B-Instruct · 87.2 pass@1 (%) · zoomed horizontal axisDeepSeek-V3 (Nov 2024) · 86.6 pass@1 (%) · zoomed horizontal axis86.2489.36ZOOMED · NOT ZERO
pass@1 (%)查看图表
EvalPlus2026-10-23

MBPP+ 代码测试

#1O1 Preview (Sept 2024)80.2
MBPP+ code tests. Zoomed horizontal axis, from 73.76 to 81.04O1 Preview (Sept 2024) · 80.2 pass@1 (%) · zoomed horizontal axisO1 Mini (Sept 2024) · 78.8 pass@1 (%) · zoomed horizontal axisQwen2.5-Coder-32B-Instruct · 77 pass@1 (%) · zoomed horizontal axisDeepSeek-Coder-V2-Instruct · 75.1 pass@1 (%) · zoomed horizontal axisGemini 1.5 Pro 002 · 74.6 pass@1 (%) · zoomed horizontal axis73.7681.04ZOOMED · NOT ZERO
pass@1 (%)查看图表
数据说明

数据可信，来源可查。

我们使用下列三个评测发布方的开放结果重绘图表，保留原始评分口径与数据日期。单项任务第一名不代表综合第一名。

ArenaCC BY 4.0 · 官方来源 ↗MTEBCC0 · 官方来源 ↗EvalPlusApache-2.0 · 官方来源 ↗
其他评测机构与方法资料

以下仅是评测方法参考链接。未取得可核验数据和适用授权的结果，不会复制到本站。

HELM查看发布方 ↗Artificial Analysis查看发布方 ↗SWE-bench查看发布方 ↗LiveCodeBench查看发布方 ↗WMT查看发布方 ↗MathArena查看发布方 ↗GPQA查看发布方 ↗BFCL查看发布方 ↗