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。 排序方式: 失败次数 ↑.
313/313
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 | 分数 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #186 | Claude Sonnet 5 none | Anthropic | 8 | 6.1 | $0.548 | 7/22 | 6.05s |
| #187 | Gemini 3.1 Flash Lite minimal | 8 | 6.1 | $0.047 | 10/22 | 1.85s | |
| #193 | Inkling low | Thinkingmachines | 8 | 6.1 | $0.187 | 10/22 | 5.11s |
| #194 | Trinity Large Thinking medium | Arcee AI | 8 | 6.1 | $0.756 | 8/22 | 85.4s |
| #211 | GLM 5.3 Flash low | Z.ai | 8 | 5.9 | $0.015 | 10/22 | 9.65s |
| #267 | DeepSeek V3.2 none | DeepSeek | 8 | 5.0 | $0.054 | 6/22 | 17.9s |
| #300 | Hy3 preview none | Tencent | 8 | 4.0 | $0.007 | 4/21 | 12.9s |
| #311 | Laguna Xs.2 none | Poolside | 8 | 3.8 | $0.004 | 5/19 | 806ms |
| #312 | gpt-oss-120b none | OpenAI | 8 | 3.7 | $0.010 | 6/19 | 21.6s |
| #108 | GPT-5.6 Luna medium | OpenAI | 9 | 7.4 | $0.072 | 13/22 | 7.43s |
| #118 | Qwen3.7 Plus none | Qwen | 9 | 7.3 | $0.106 | 12/22 | 12.1s |
| #126 | Muse Glimmer 30B medium | Meta | 9 | 7.2 | $0.211 | 11/22 | 30.6s |
| #129 | Step 3.7 Flash low | Stepfun | 9 | 7.1 | $0.390 | 11/22 | 17.9s |
| #132 | GPT-5.6 Sol none | OpenAI | 9 | 7.0 | $0.201 | 12/22 | 2.17s |
| #140 | KAT-Coder-Pro V2.5 medium | Kwaipilot | 9 | 6.9 | $0.478 | 11/22 | 24.9s |