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: 分数 ↑.
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） |
|---|---|---|---|---|---|---|---|
| #206 | Space Bunny Alpha max | Stealth | 2 | 0.7 | $0.000 | 11/23 | 36.2s |
| #204 | Mercury 2.5 Preview medium | Inception | 1 | 0.7 | $0.039 | 15/23 | 6.66s |
| #203 | Inkling low | Thinkingmachines | 2 | 0.7 | $0.293 | 11/23 | 6.81s |
| #202 | Qwen3.5-Flash medium | Qwen | 1 | 0.7 | $0.184 | 12/23 | 83.7s |
| #201 | Kimi K2.6 medium | Moonshot AI | 2 | 0.7 | $1.594 | 12/23 | 121.3s |
| #200 | Claude Opus 4.8 none | Anthropic | 1 | 0.7 | $2.334 | 13/23 | 7.54s |
| #199 | Claude Sonnet 5.5 low | Anthropic | 5 | 0.7 | $0.909 | 10/23 | 7.46s |
| #198 | GPT-5.6 Luna low | OpenAI | 1 | 0.7 | $0.049 | 11/23 | 6.96s |
| #195 | Gemini 3.1 Flash Lite medium | 1 | 0.7 | $0.153 | 13/23 | 6.27s | |
| #194 | Minimax M3.1 Flash max | Minimax | 1 | 0.7 | $0.000 | 11/23 | 40.9s |
| #193 | Muse Glimmer 30B low | Meta | 3 | 0.7 | $0.219 | 11/23 | 22.4s |
| #189 | Step 5 Preview high | Stepfun | 1 | 0.7 | $4.513 | 10/23 | 154.4s |
| #187 | Kimi K2.7 Code medium | Moonshot AI | 1 | 0.7 | $1.136 | 11/23 | 81.2s |
| #186 | Gemini 3.1 Flash Lite Preview medium | 1 | 0.7 | $0.158 | 14/23 | 6.78s | |
| #184 | GPT-5.2 Chat default | OpenAI | 1 | 0.7 | $0.594 | 13/23 | 7.43s |