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 ↑.
293/293
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） |
|---|---|---|---|---|---|---|---|
| #24 | Claude Opus 5.5 low | Anthropic | 1 | 0.9 | $1.733 | 19/23 | 25.7s |
| #25 | Claude Opus 5.5 high | Anthropic | 1 | 0.9 | $2.133 | 19/23 | 27.6s |
| #30 | Muse Spark Open high | Meta | 1 | 0.9 | $1.782 | 17/23 | 152.0s |
| #36 | DeepSeek V4.1 Flash 1017 low | DeepSeek | 1 | 0.9 | $0.355 | 16/23 | 32.5s |
| #37 | Qwen3.8 Max (0902) high | Qwen | 1 | 0.9 | $3.306 | 18/23 | 186.1s |
| #39 | Claude Haiku 5.5 xhigh | Anthropic | 1 | 0.9 | $0.140 | 18/23 | 16.6s |
| #43 | Seed 2.1 Turbo high | Bytedance Seed | 1 | 0.9 | $1.892 | 17/23 | 170.5s |
| #53 | GLM 5.3 Flash max | Z.ai | 1 | 0.9 | $0.089 | 17/23 | 54.0s |
| #59 | Pareto 26.9 default | Unbiased | 1 | 0.9 | $1.864 | 15/23 | 28.3s |
| #60 | Qwen3.8 MAX low | Qwen | 1 | 0.9 | $1.089 | 17/23 | 25.1s |
| #63 | DeepSeek V4.1 Flash high | DeepSeek | 1 | 0.9 | $0.617 | 16/23 | 33.3s |
| #64 | DeepSeek V4.1 Flash medium | DeepSeek | 1 | 0.9 | $0.591 | 16/23 | 32.9s |
| #68 | MiMo-V2.6-Pro medium | Xiaomi | 1 | 0.9 | $0.586 | 14/23 | 148.1s |
| #69 | Seed 2.1 Turbo medium | Bytedance Seed | 1 | 0.9 | $2.042 | 16/23 | 189.1s |
| #71 | DeepSeek V4.1 Flash max | DeepSeek | 1 | 0.9 | $1.204 | 15/23 | 61.6s |