| 能力 | 选中模型 | P值 | U/R/C | 降级链 |
|---|---|---|---|---|
| 文本对话 | volcengine/doubao-pro | 0.9259 | U=0.93 R=0.10 C=0.04 | agnes/omni-free → ollama/qwen2.5-7b → siliconflow/qwen2.5-72b |
| 代码生成 | volcengine/doubao-pro | 0.9259 | U=0.93 R=0.10 C=0.04 | ollama/llama3.1-8b → ollama/qwen2.5-7b → siliconflow/qwen2.5-72b |
| 视频生成 | volcengine/seedance-2.0-mini | 0.9168 | U=0.93 R=0.11 C=0.40 | aliyun/wan3.0-video |
| 图像生成 | agnes/omni-free | 0.9167 | U=0.84 R=0.11 C=0.08 | — |
| 向量嵌入 | ollama/qwen2.5-7b | 0.9173 | U=0.84 R=0.03 C=0.02 | — |
| RAG推理 | siliconflow/deepseek-v3 | 0.9172 | U=0.96 R=0.11 C=0.06 | kimi/k2 → aliyun/qwen-max |
| 函数调用 | volcengine/doubao-pro | 0.9214 | U=0.91 R=0.10 C=0.04 | aliyun/qwen-max → zhipu/glm-4-flash |
| 长文本 | volcengine/doubao-pro | 0.9214 | U=0.91 R=0.10 C=0.04 | kimi/k2 → siliconflow/qwen2.5-72b → siliconflow/deepseek-v3 |