未遵循指令 失败
看看哪些 AI 模型最常遇到 未遵循指令，让你在选择前先发现稳定性风险。 排序方式: 失败次数 ↑.
140/140
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 | 未遵循指令 次数 | 分数 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #9 | Gemini 3.5 Flash medium | 1 | 9.1 | $0.642 | 19/22 | 8.20s | |
| #12 | Grok 4.5 high | X AI | 1 | 8.9 | $1.707 | 17/22 | 76.5s |
| #23 | Claude Sonnet 5 medium | Anthropic | 1 | 8.3 | $0.922 | 16/22 | 12.5s |
| #25 | Gemini 2.5 Flash medium | 1 | 8.2 | $0.643 | 15/22 | 21.2s | |
| #28 | Inkling high | Thinkingmachines | 1 | 8.0 | $1.006 | 15/22 | 64.2s |
| #29 | Step 3.7 Flash medium | Stepfun | 1 | 8.0 | $0.515 | 14/22 | 26.4s |
| #30 | GPT-5.2 Chat none | OpenAI | 1 | 8.0 | $0.604 | 14/22 | 7.65s |
| #31 | GLM 5.2 high | Z.ai | 1 | 8.0 | $0.970 | 14/22 | 62.7s |
| #37 | Qwen3.6 Plus medium | Qwen | 1 | 7.8 | $0.405 | 15/22 | 43.1s |
| #42 | GLM 5 medium | Z.ai | 1 | 7.7 | $0.307 | 15/21 | 33.5s |
| #43 | Claude Opus 4.6 medium | Anthropic | 1 | 7.7 | $3.059 | 13/22 | 34.3s |
| #49 | GLM 5 Turbo medium | Z.ai | 1 | 7.6 | $0.323 | 14/21 | 23.0s |
| #52 | Kimi K2.7 Code medium | Moonshot AI | 1 | 7.5 | $0.751 | 12/22 | 84.2s |
| #63 | Claude Sonnet 4.6 none | Anthropic | 1 | 7.3 | $0.661 | 12/22 | 8.12s |
| #64 | Gemini 3.1 Flash Lite Preview medium | 1 | 7.3 | $0.115 | 13/22 | 4.61s |