AI BENCHY 分类失败
编程：答案错误
编程
答案错误
看看哪些 AI 模型在 编程 上最容易遇到 答案错误，更快找出薄弱点。
134/134
筛选模型
没有模型匹配当前搜索和筛选条件。
| 排名 | 模型 | 公司 | 答案错误 次数 | 分类得分 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #152 | Mistral Small 4 medium | Mistral | 3 | 4.4 | $0.068 | 0/3 | 40.0s |
| #154 | Qwen3.5-9B none | Qwen | 3 | 3.9 | $0.006 | 0/3 | 5.60s |
| #158 | GPT-4o-mini none | OpenAI | 3 | 3.2 | $0.006 | 0/3 | 1.63s |
| #160 | GLM 4.7 Flash none | Z.ai | 3 | 4.3 | $0.004 | 0/3 | 2.54s |
| #161 | Nemotron 3 Super none | NVIDIA | 3 | 3.3 | $0.006 | 0/3 | 2.64s |
| #164 | GPT-5.4 Nano none | OpenAI | 3 | 4.6 | $0.011 | 0/3 | 2.22s |
| #170 | Mercury 2 none | Inception | 3 | 3.4 | $0.011 | 0/3 | 1.03s |
| #34 | Qwen3.5-27B medium | Qwen | 2 | 6.2 | $0.536 | 1/3 | 160.7s |
| #37 | GPT-5.6 Terra medium | OpenAI | 2 | 6.1 | $0.496 | 1/3 | 7.19s |
| #40 | Gemini 3.1 Flash Lite Preview medium | 2 | 5.5 | $0.068 | 1/3 | 4.09s | |
| #41 | Qwen3.5 Plus 2026-04-20 medium | Qwen | 2 | 6.2 | $0.317 | 1/3 | 125.3s |
| #42 | Gemini 3.1 Flash Lite medium | 2 | 5.5 | $0.071 | 1/3 | 3.81s | |
| #47 | GPT-5.6 Terra low | OpenAI | 2 | 6.6 | $0.343 | 1/3 | 9.56s |
| #50 | GPT-5.6 Luna high | OpenAI | 2 | 5.5 | $0.924 | 1/3 | 15.6s |
| #54 | GPT-5.6 Luna medium | OpenAI | 2 | 5.4 | $0.258 | 1/3 | 10.4s |