通用智能：未遵循指令
通用智能
未遵循指令
看看哪些 AI 模型在 通用智能 上最容易遇到 未遵循指令，更快找出薄弱点。 排序方式: 失败次数 ↑.
142/142
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 | 分类得分 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #34 | Claude Sonnet 5.5 max | Anthropic | 1 | 0.5 | $8.557 | 0/1 | 56.7s |
| #37 | Qwen3.8 Max (0902) high | Qwen | 1 | 0.6 | $3.306 | 0/1 | 75.1s |
| #38 | GPT-5.3-Codex medium | OpenAI | 1 | 0.5 | $1.318 | 0/1 | 4.87s |
| #43 | Seed 2.1 Turbo high | Bytedance Seed | 1 | 0.6 | $1.892 | 0/1 | 25.7s |
| #56 | Muse Spark 1.2 medium | Meta | 1 | 0.6 | $1.590 | 0/1 | 10.7s |
| #59 | Pareto 26.9 default | Unbiased | 1 | 0.6 | $1.864 | 0/1 | 6.55s |
| #60 | Qwen3.8 MAX low | Qwen | 1 | 0.6 | $1.089 | 0/1 | 5.29s |
| #61 | GPT-5.4 medium | OpenAI | 1 | 0.5 | $2.213 | 0/1 | 4.92s |
| #65 | Muse Spark 1.2 low | Meta | 1 | 0.5 | $0.881 | 0/1 | 8.46s |
| #66 | GPT-5.2 medium | OpenAI | 1 | 0.4 | $1.558 | 0/1 | 4.32s |
| #67 | Qwen3.8 27B high | Qwen | 1 | 0.5 | ~$0.298 | 0/1 | 8.45s |
| #69 | Seed 2.1 Turbo medium | Bytedance Seed | 1 | 0.6 | $2.042 | 0/1 | 40.6s |
| #73 | Claude Sonnet 5.5 xhigh | Anthropic | 1 | 0.6 | $1.516 | 0/1 | 9.10s |
| #79 | Claude Sonnet 5 medium | Anthropic | 1 | 0.5 | $1.438 | 0/1 | 4.32s |
| #82 | Nemotron 3.5 Super VL high | NVIDIA | 1 | 0.6 | $0.349 | 0/1 | 14.3s |