AI BENCHY 失败分析
答案错误 失败
看看哪些 AI 模型最常遇到 答案错误，让你在选择前先发现稳定性风险。
| 排名 | 模型 | 公司 | 答案错误 次数 | 分数 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|
| #85 | Kimi K2.5 medium | Moonshot AI | 5 | 6.7 | 9/20 | 89.0s |
| #102 | Nemotron 3 Super medium | NVIDIA | 5 | 5.9 | 8/20 | 20.9s |
| #11 | GPT-5.5 medium | OpenAI | 4 | 8.7 | 16/20 | 37.9s |
| #13 | Gemini 3 Flash Preview low | 4 | 8.6 | 16/20 | 5.81s | |
| #17 | GPT-5.3-Codex medium | OpenAI | 4 | 8.3 | 14/20 | 16.0s |
| #23 | Qwen3.5 Plus 2026-02-15 medium | Qwen | 4 | 8.1 | 14/20 | 67.6s |
| #26 | Hy3 preview low | Tencent | 4 | 7.9 | 15/20 | 24.6s |
| #27 | Gemini 3.5 Flash minimal | 4 | 7.9 | 14/20 | 1.58s | |
| #31 | GLM 5 Turbo medium | Z.ai | 4 | 7.9 | 13/20 | 22.7s |
| #35 | Qwen3.6 35B A3B medium | Qwen | 4 | 7.8 | 14/20 | 17.3s |
| #36 | Grok 4.3 medium | X AI | 4 | 7.8 | 13/20 | 49.2s |
| #45 | Grok Build 0.1 medium | X AI | 4 | 7.7 | 13/20 | 42.4s |
| #56 | GLM 5.1 medium | Z.ai | 4 | 7.4 | 12/20 | 33.5s |
| #58 | MiMo-V2.5 medium | Xiaomi | 4 | 7.4 | 12/20 | 20.3s |
| #61 | Step 3.5 Flash medium | Stepfun | 4 | 7.4 | 11/19 | 43.2s |