多模态模型
接受图像输入的开源模型越来越多。本页列出语料中的全部多模态模型，以及识别它们的 config 证据。
判定标准（客观、机械）：随权重发布的 config.json 中含视觉相关字段——图像/视觉 token id、视觉编码器设置或图像分块字段。不作其他推断。
| 模型 | 实验室 | 发布时间 | 总参数 | MoE | 注意力 | config 证据 |
|---|---|---|---|---|---|---|
| Step-3 | StepFun | 2025-07 | 321 B | 稀疏 MoE | MHA | im_end_token im_patch_token im_start_token +3 |
| Hunyuan 7B | Hunyuan | 2025-08 | 7.8 B | 稠密 | GQA | im_end_id im_newline_id im_start_id |
| Kimi K2.5 | Kimi | 2026-02 | 1,040.2 B | 稀疏 MoE | MLA | use_unified_vision_chunk |
| Qwen3.5 122B A10B | Qwen | 2026-02 | 119.9 B | 稀疏 MoE | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.5 27B | Qwen | 2026-02 | 23.7 B | 稠密 | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.5 35B A3B | Qwen | 2026-02 | 33.9 B | 稀疏 MoE | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.5 397B A17B | Qwen | 2026-02 | 392.6 B | 稀疏 MoE | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.5 9B | Qwen | 2026-02 | 7.9 B | 稠密 | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.5 9B Base | Qwen | 2026-02 | 7.9 B | 稠密 | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Kimi K2.6 | Kimi | 2026-04 | 1,040.2 B | 稀疏 MoE | MLA | use_unified_vision_chunk |
| Qwen3.6 27B | Qwen | 2026-04 | 23.7 B | 稠密 | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.6 35B A3B | Qwen | 2026-04 | 33.9 B | 稀疏 MoE | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Hunyuan-TurboS 7B | Hunyuan | 2026-05 | 7.8 B | 稠密 | GQA | im_end_id im_newline_id im_start_id |
| Step-3.7-Flash | StepFun | 2026-05 | 198 B | 稀疏 MoE | MHA | im_end_token im_patch_token im_start_token +5 |
| MiniMax-M3 | MiniMax | 2026-06 | 443.4 B | 稀疏 MoE | GQA | image_grid_pinpoints image_seq_length image_token_index +3 |
| MiniMax-M3 MXFP8 | MiniMax | 2026-06 | 443.4 B | 稀疏 MoE | GQA | image_grid_pinpoints image_seq_length image_token_index +3 |
| Kimi K3 | Kimi | 2026-07 | 2,780 B | 稀疏 MoE | MLA | image_placeholder |
| GLM-5.3-Flash | GLM | 2026-08 | 328.9 B | 稀疏 MoE | MLA | image_token_id image_start_token_id image_end_token_id |
| Qwen3.8 27B | Qwen | 2026-08 | 23.7 B | 稠密 | GQA | image_token_id vision_end_token_id vision_start_token_id |
| Qwen3.8-Flash-Next | Qwen | 2026-08 | 123.6 B | 稀疏 MoE | GQA | image_token_id vision_end_token_id vision_start_token_id |
完整字段列表见各模型页；表中仅展示前几个字段作为证据。