AI 模型评测榜：推理、Agentic 编程与视觉

名次前 10

1. 1 GPT-6 Astra (xhigh)OpenAI 86.53
2. 2 Opus 5.5 (xhigh ~Estimated)Anthropic Claude 77.69
3. 3 Fable 5.1 (xhigh ~Estimated)Anthropic Claude 70.47
4. 4 Gemini 4 Argon (high)Google Gemini 69.13
5. 5 GPT-6.1 Sol (xhigh)OpenAI 67.79
6. 6 Opus 5.5 (xhigh w/ refuse)Anthropic Claude 66.63
7. 7 Kimi-K3.1-Pro (max)月之暗面 Kimi 61.65
8. 8 Muse Wonder 1.0 (xhigh)Meta 59.64
9. 9 Fable 5.1 (xhigh w/ refuse)Anthropic Claude 58.69
10. 10 Sonnet 5.5 (xhigh)Anthropic Claude 57.41

名次前 10

1. 1 Opus 5.5 (max)Anthropic Claude · Claude Code
2. 2 GPT-6 Astra (max)OpenAI · Codex CLI
3. 3 GPT-6.1 Sol (max)OpenAI · Codex CLI
4. 4 Fable 5 (high)Anthropic Claude · Claude Code
5. 5 Gemini 4 Argon (high)Google Gemini · OpenCode CLI
6. 6 DeepSeek V4.1 Flash 1017 (max)DeepSeek 深度求索 · DSH
7. 7 Kimi-K3.1-Pro (max)月之暗面 Kimi · Kimi Code
8. 8 Kimi-K3 (max)月之暗面 Kimi · Claude Code
9. 9 Grok 4.7 (high)xAI Grok · Grok Build
10. 10 GLM-5.3 (max)Z.AI（智谱国际版） · Claude Code

名次前 10

1. 1 Gemini 3 ProGoogle Gemini 70.47
2. 2 Gemini 2.5 ProGoogle Gemini 61.80
3. 3 GPT-5 (high)OpenAI 53.79
4. 4 Seed-1.6-vision (Think)字节豆包（火山方舟） 55.85
5. 5 qwen3-vl-235b-a22b (Think)阿里云百炼（通义千问） 53.50
6. 6 Seed-1.6-thinking-0715七牛云 AI 50.87
7. 7 GLM-4.5VZ.AI（智谱国际版） 49.42
8. 8 Seed-1.6-vision字节豆包（火山方舟） 43.74
9. 9 Step-3阶跃星辰 StepFun 45.34
10. 10 Qwen3-vl-235b-a22b阿里云百炼（通义千问） 45.62