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 ↑.
443/443
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） |
|---|---|---|---|---|---|---|---|
| #1 | Gemini 3.6 Flash high | 1 | 1.0 | $0.900 | 22/23 | 18.9s | |
| #3 | GPT-6 Sol high | OpenAI | 1 | 1.0 | $0.921 | 22/23 | 23.6s |
| #6 | GPT-6.1 Sol xhigh | OpenAI | 1 | 1.0 | $1.377 | 22/23 | 33.1s |
| #7 | GPT-6 Astra high | OpenAI | 1 | 1.0 | $4.458 | 22/23 | 20.2s |
| #8 | GPT-6 Astra xhigh | OpenAI | 1 | 1.0 | $6.213 | 22/23 | 29.7s |
| #17 | Gemini 4 Argon high | 1 | 1.0 | $1.802 | 21/23 | 19.0s | |
| #58 | Claude Fable 5 medium | Anthropic | 1 | 0.9 | $5.089 | 18/23 | 16.8s |
| #81 | Claude Haiku 5.5 max | Anthropic | 1 | 0.8 | $0.509 | 19/23 | 61.7s |
| #199 | Claude Sonnet 5.5 low | Anthropic | 1 | 0.7 | $0.909 | 10/23 | 7.46s |
| #227 | Claude Sonnet 5.5 medium | Anthropic | 1 | 0.7 | $1.100 | 9/23 | 8.03s |
| #417 | Command A+ medium | Cohere | 1 | 0.3 | $0.064 | 0/23 | 91.7s |
| #423 | Command A+ high | Cohere | 1 | 0.3 | $0.079 | 0/23 | 90.0s |
| #427 | Command A+ low | Cohere | 1 | 0.3 | $0.073 | 0/23 | 92.5s |
| #442 | Step 3.5 Flash none | Stepfun | 1 | 0.2 | $0.020 | 6/12 | 39.0s |
| #9 | Gemini 3.7 Flash high | 2 | 1.0 | $0.528 | 21/23 | 15.1s |