无效工具调用 失败
看看哪些 AI 模型最常遇到 无效工具调用，让你在选择前先发现稳定性风险。 排序方式: 失败次数 ↑.
190/190
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 | 无效工具调用 次数 | 分数 | 总成本 | 测试正确 | 响应时间（平均） |
|---|---|---|---|---|---|---|---|
| #75 | Qwen3.8 Max (0902) low | Qwen | 1 | 0.9 | $1.131 | 18/23 | 29.1s |
| #77 | Muse Spark 1.3 low | Meta | 1 | 0.9 | $0.911 | 16/23 | 17.7s |
| #78 | Claude Sonnet 5 medium | Anthropic | 1 | 0.9 | $1.438 | 16/23 | 15.0s |
| #79 | Qwen3.7 Max medium | Qwen | 1 | 0.9 | $1.689 | 17/23 | 48.4s |
| #82 | Qwen3.8 2.4T A95B low | Qwen | 1 | 0.8 | $4.016 | 16/23 | 120.4s |
| #84 | Muse Glimmer 30B xhigh | Meta | 1 | 0.8 | $0.585 | 14/23 | 97.7s |
| #90 | GPT-5.6 Terra high | OpenAI | 1 | 0.8 | $0.698 | 15/23 | 13.6s |
| #96 | Ling 3.1 Flash medium | Inclusionai | 1 | 0.8 | $0.000 | 15/23 | 88.7s |
| #97 | Qwen3.7 Plus medium | Qwen | 1 | 0.8 | $0.415 | 16/23 | 59.7s |
| #105 | Claude Opus 4.7 medium | Anthropic | 1 | 0.8 | $3.636 | 18/23 | 14.1s |
| #106 | DeepSeek V4 Flash 0423 high | DeepSeek | 1 | 0.8 | $0.261 | 14/23 | 56.0s |
| #109 | Nemotron 3 Ultra medium | NVIDIA | 1 | 0.8 | $0.674 | 14/23 | 40.3s |
| #111 | Gemini 3.8 Flash medium | 1 | 0.8 | $0.769 | 19/23 | 16.6s | |
| #112 | Qwen3.8 Flash Next low | Qwen | 1 | 0.8 | ~$0.045 | 15/23 | 23.7s |
| #113 | GPT-5.6 Terra low | OpenAI | 1 | 0.8 | $0.435 | 14/23 | 7.58s |