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） |
|---|---|---|---|---|---|---|
| #160 | LFM2-24B-A2B none | Liquid | 9 | 4.2 | 2/16 | 782ms |
| #162 | Nemotron 3 Nano Omni 30b A3b Reasoning none | NVIDIA | 9 | 4.1 | 2/19 | 728ms |
| #36 | Qwen3.5 Plus 2026-04-20 medium | Qwen | 8 | 7.6 | 13/21 | 46.4s |
| #39 | Qwen3.6 Flash medium | Qwen | 8 | 7.5 | 12/21 | 19.2s |
| #48 | Gemini 3 Flash Preview none | 8 | 7.4 | 13/21 | 1.65s | |
| #57 | Step 3.7 Flash low | Stepfun | 8 | 7.3 | 12/21 | 15.7s |
| #70 | GPT-5.4 Nano medium | OpenAI | 8 | 7.0 | 11/21 | 12.0s |
| #81 | Mercury 2 medium | Inception | 8 | 6.6 | 10/21 | 2.24s |
| #85 | Gemma 4 31B none | 8 | 6.5 | 10/21 | 4.05s | |
| #87 | Gemini 3.1 Flash Lite minimal | 8 | 6.4 | 10/21 | 1.33s | |
| #126 | gpt-oss-120b none | OpenAI | 8 | 5.4 | 6/19 | 21.6s |
| #146 | Laguna Xs.2 none | Poolside | 8 | 4.8 | 5/19 | 806ms |
| #156 | Hy3 preview none | Tencent | 8 | 4.4 | 4/21 | 12.9s |
| #34 | Qwen3.7 Max none | Qwen | 7 | 7.7 | 14/21 | 1.30s |
| #40 | Gemini 3.1 Flash Lite Preview medium | 7 | 7.5 | 13/21 | 3.96s |