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） |
|---|---|---|---|---|---|---|---|
| #73 | Claude Sonnet 5.5 xhigh | Anthropic | 1 | 0.9 | $1.516 | 16/23 | 15.0s |
| #78 | Muse Spark 1.3 low | Meta | 1 | 0.9 | $0.911 | 16/23 | 17.7s |
| #79 | Claude Sonnet 5 medium | Anthropic | 1 | 0.9 | $1.438 | 16/23 | 15.0s |
| #85 | Muse Glimmer 30B xhigh | Meta | 1 | 0.8 | $0.585 | 14/23 | 97.7s |
| #87 | Inkling high | Thinkingmachines | 1 | 0.8 | $1.238 | 16/23 | 62.7s |
| #90 | GLM 5.2 high | Z.ai | 1 | 0.8 | $4.632 | 15/23 | 69.1s |
| #92 | Claude Haiku 5.5 high | Anthropic | 1 | 0.8 | $0.088 | 15/23 | 10.4s |
| #95 | GPT-6 Luna medium | OpenAI | 1 | 0.8 | $0.047 | 15/23 | 18.1s |
| #99 | Gemini 4 Argon low | 1 | 0.8 | $0.438 | 15/23 | 5.72s | |
| #101 | Grok 4.5 high | X AI | 1 | 0.8 | $2.687 | 17/23 | 72.5s |
| #105 | Seed-2.0-Code low | Bytedance Seed | 1 | 0.8 | $0.881 | 14/23 | 73.0s |
| #114 | Qwen3.8 Flash Next low | Qwen | 1 | 0.8 | ~$0.045 | 15/23 | 23.7s |
| #117 | GLM 5.3 FlashX max | Z.ai | 1 | 0.8 | $0.623 | 15/23 | 44.0s |
| #121 | Gemini 3.5 Flash medium | 1 | 0.8 | $1.316 | 18/23 | 11.8s | |
| #123 | MiMo-V2.6-Flash medium | Xiaomi | 1 | 0.8 | $0.218 | 12/23 | 61.9s |