答案错误 失败
看看哪些 AI 模型最常遇到 答案错误，让你在选择前先发现稳定性风险。 排序方式: 失败次数 ↑.
313/313
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 | 答案错误 次数 | 分数 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #36 | Claude Opus 4.8 medium | Anthropic | 4 | 8.8 | $1.983 | 17/22 | 13.0s |
| #37 | Claude Opus 5 low | Anthropic | 4 | 8.7 | $1.121 | 18/22 | 6.99s |
| #38 | Gemini 3.6 Flash low | 4 | 8.7 | $0.251 | 18/22 | 4.92s | |
| #41 | Muse Spark 1.2 high | Meta | 4 | 8.7 | $1.771 | 14/22 | 26.6s |
| #43 | Muse Spark 1.1 medium | Meta | 4 | 8.6 | $1.420 | 15/22 | 26.2s |
| #45 | Qwen3.6 Max Preview medium | Qwen | 4 | 8.6 | $1.132 | 17/22 | 66.5s |
| #49 | Muse Spark 1.2 low | Meta | 4 | 8.4 | $0.655 | 15/22 | 9.77s |
| #59 | Muse Spark 1.3 low | Meta | 4 | 8.2 | $0.689 | 15/22 | 16.0s |
| #61 | Qwen3.8 2.4T A95B low | Qwen | 4 | 8.1 | $2.672 | 15/22 | 119.0s |
| #63 | Muse Glimmer 30B xhigh | Meta | 4 | 8.1 | $0.471 | 13/22 | 95.6s |
| #64 | Inkling high | Thinkingmachines | 4 | 8.0 | $1.046 | 15/22 | 62.6s |
| #67 | DeepSeek V4 Flash 0731 high | DeepSeek | 4 | 8.0 | $0.084 | 14/22 | 114.9s |
| #69 | Inkling medium | Thinkingmachines | 4 | 8.0 | $0.383 | 15/22 | 15.7s |
| #74 | Qwen3.8 27B low | Qwen | 4 | 7.9 | ~$0.071 | 15/22 | 39.1s |
| #80 | Claude Sonnet 4.6 medium | Anthropic | 4 | 7.8 | $2.126 | 14/22 | 28.1s |