Arena Agent · Bash Reliability
How often the model’s bash commands succeed. Errors outside its control aren’t counted. E.g. network issues, permission denied, etc. 完整榜单 53 个模型,样本量 9,893,691 次观测。
综合总榜535 个行为信号加权后的总排名任务确认完成53用户在会话里明确确认「任务完成了」的比例好评 vs 差评53用户明确表扬多于明确抱怨的程度可操控性53用户纠正、反驳之后,模型能改到点子上的比例Bash Reliability53How often the model’s bash commands succeed. Errors outside its control aren’t counted. E.g. network issues, permission denied, etc.工具幻觉53调用根本不存在的工具的情况,分越高越少乱编
| Rank | Model | 信号分 |
|---|---|---|
1 | Claude Opus 4.8 (High) | +15.8 |
2 | Claude Sonnet 5 (High) | +15.7 |
3 | Claude Opus 5.5 (Max) | +15.4 |
4 | GPT 5.5 (xHigh) | +13.3 |
5 | Claude Sonnet 5.5 (Max) | +13.2 |
6 | Claude Fable 5 (High) | +12.9 |
7 | GPT 5.6 Sol (xHigh) | +12.0 |
8 | Claude Opus 5 (Max) | +11.9 |
9 | Gemini 3.8 Flash (High) | +11.3 |
10 | GPT 6 Sol (Max) | +11.0 |
11 | GPT 6 Astra (Max) | +10.9 |
12 | GPT 6 Luna (Max) | +10.9 |
13 | Claude Fable 5.1 (Max) | +10.8 |
14 | GPT 5.4 (High) | +10.3 |
15 | Grok 4.6 (xHigh) | +9.3 |
16 | Grok 4.7 (xHigh) | +9.0 |
17 | Gemini 3.7 Flash (High) | +7.7 |
18 | Claude Haiku 5.5 (High) | +7.7 |
19 | GPT 5.6 Luna (xHigh) | +7.4 |
20 | GPT 5.5 | +7.0 |
21 | Qwen3.7 Max | +6.0 |
22 | Inkling Small | +4.8 |
23 | MiMo V2.6 Pro | +4.5 |
24 | Gemini 3.1 Pro Preview | +3.2 |
25 | Inkling | +2.8 |
26 | GPT 6.1 Sol (Max) | +2.6 |
27 | Muse Spark 1.3 (Max) | +2.5 |
28 | DeepSeek V4 Pro (High) (0813) | +2.4 |
29 | GPT 5.6 Terra (xHigh) | +2.3 |
30 | Minimax M3 | +2.2 |
31 | Mistral Large 4 | +2.0 |
32 | Qwen3.8 Max | +1.3 |
33 | Mimo V2.5 Pro | +0.6 |
34 | Gemini 3.6 Flash (High) | +0.5 |
35 | Deepseek V4.1 Flash (Max) | -0.4 |
36 | Qwen3.7 Plus | -0.5 |
37 | Grok 4.5 | -1.2 |
38 | MiMo V2.6 Flash | -1.8 |
39 | Kimi K3 (Max) | -1.9 |
40 | Gemini 4 Argon (High) | -1.9 |
41 | Hy4 preview | -2.0 |
42 | Muse Spark 1.1 | -2.2 |
43 | GLM 5.2 (Max) | -2.9 |
44 | GLM 5.3 (Max) | -3.8 |
45 | Muse Spark 1.2 (xHigh) | -4.1 |
46 | Nemotron 3 Ultra | -7.3 |
47 | Mistral Medium 3.5 | -8.1 |
48 | Qwen 3.8 27B | -10.0 |
49 | Step 5 Preview | -10.1 |
50 | GLM 5.3 Flash | -10.2 |
51 | Hy3 | -14.7 |
52 | Qwen3.8 Flash Next | -18.5 |
53 | Solar Pro 4 | -51.6 |
数据更新: 2026-10-14数据来源: Arena AI ↗