AI BENCHY 失败分析
额外格式 失败
看看哪些 AI 模型最常遇到 额外格式，让你在选择前先发现稳定性风险。 排序方式: 失败次数 ↑.
| 排名 | 模型 | 公司 | 额外格式 次数 | 分数 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|
| #30 | Qwen3.5-27B medium | Qwen | 1 | 7.8 | 13/21 | 68.4s |
| #38 | Grok 4.3 medium | X AI | 1 | 7.6 | 13/21 | 47.5s |
| #51 | Mimo V2 PRO medium | Xiaomi | 1 | 7.4 | 12/21 | 22.2s |
| #55 | GLM 5.1 medium | Z.ai | 1 | 7.3 | 12/21 | 33.7s |
| #64 | MiMo-V2-Flash medium | Xiaomi | 1 | 7.2 | 12/21 | 20.1s |
| #65 | Grok 4.20 medium | X AI | 1 | 7.1 | 12/21 | 27.7s |
| #67 | MiniMax M3 medium | Minimax | 1 | 7.1 | 11/21 | 68.2s |
| #79 | Hunter Alpha medium | OpenRouter | 1 | 6.7 | 8/18 | 10.3s |
| #80 | Mimo V2 Omni medium | Xiaomi | 1 | 6.7 | 10/21 | 41.2s |
| #101 | Mimo V2 Omni none | Xiaomi | 1 | 6.0 | 8/21 | 2.44s |
| #113 | DeepSeek V4 Pro none | DeepSeek | 1 | 5.7 | 7/21 | 12.4s |
| #121 | Owl Alpha none | Openrouter | 1 | 5.5 | 7/21 | 9.88s |
| #127 | Grok 4.20 none | X AI | 1 | 5.4 | 6/18 | 1.11s |
| #140 | Qwen3 Coder Next none | Qwen | 1 | 4.9 | 5/21 | 8.62s |
| #143 | MiMo-V2.5 none | Xiaomi | 1 | 4.9 | 5/21 | 2.20s |