MiniCPM-V — OpenBMB 的本地 AI 模型
作者： Jakub Rusinowski · 最后更新：
OpenBMB's pocket-sized multimodal line, backed by ModelBest and Tsinghua University, built around one goal: image and video understanding that actually runs on a phone. Version 4.5 is the 8B that beat much larger proprietary models on OpenCompass; version 4.6 is the 1.3B successor that cut visual-encoding FLOPs by more than half and ships on iOS, Android and HarmonyOS.
变体
MiniCPM-V 最小的变体在 Q4_K_M 下约需 2 GB 显存——量化权重加框架开销，不含 KV 缓存。
| 模型 | Q4 下显存 | 显存 | 上下文 | 运行 |
|---|---|---|---|---|
| MiniCPM-V 4.6 → 1.3B | ~1.6 GB | 262,144 | ollama run minicpm-v4.6 | |
| MiniCPM-V 4.5 → 8B | ~5.6 GB | 32,768 | ollama run minicpm-v4.5 |
显存为 Q4_K_M 下的量化权重加开销，与 GPU 与显存检测器使用同一引擎计算。
如何在本地运行 MiniCPM-V
安装 Ollama，然后拉取标签。
ollama run minicpm-v4.6在上方选择一个尺寸，查看它自己的显存、速度估算和安装命令。
许可证
Commercial use permitted. No usage restrictions beyond attribution.
适用于: MiniCPM-V 4.6Weights are downloadable and commercial use is permitted, subject to the licence’s acceptable-use terms.
适用于: MiniCPM-V 4.5推荐 GPU
目录中能在本地运行 MiniCPM-V（至少 2 GB 显存）的最便宜 GPU 是 Intel Arc B570 (10 GB).
我的 GPU 能运行 MiniCPM-V 吗？
- MiniCPM-V 在 AMD Radeon RX 9060 XT 8GB 上
- MiniCPM-V 在 Intel Arc B570 上
- MiniCPM-V 在 Intel Arc B580 上
- MiniCPM-V 在 NVIDIA GeForce RTX 3060 (12GB) 上
- MiniCPM-V 在 NVIDIA GeForce RTX 3070 上
- MiniCPM-V 在 NVIDIA GeForce RTX 3070 Ti 上
- MiniCPM-V 在 NVIDIA GeForce RTX 3080 (10GB) 上
- MiniCPM-V 在 NVIDIA GeForce RTX 4060 上
- MiniCPM-V 在 NVIDIA GeForce RTX 4070 上
- MiniCPM-V 在 NVIDIA GeForce RTX 4070 Super 上
- MiniCPM-V 在 NVIDIA GeForce RTX 4070 Ti 上
- MiniCPM-V 在 NVIDIA GeForce RTX 5060 上
- MiniCPM-V 在 NVIDIA GeForce RTX 5060 Ti 8GB 上
- MiniCPM-V 在 NVIDIA GeForce RTX 5070 上
MiniCPM-V — 常见问题
How much VRAM does MiniCPM-V need?
MiniCPM-V needs about 2 GB VRAM at Q4_K_M quantization for its smallest variant. Variants: MiniCPM-V 4.6 (2 GB, Q4_K_M); MiniCPM-V 4.5 (6 GB, Q4_K_M). On Apple Silicon, unified memory counts toward this requirement.
Can I run MiniCPM-V on an RTX 4090 (24 GB)?
Yes — MiniCPM-V runs on an RTX 4090 (24 GB) and other 24 GB cards such as the RTX 3090. Smaller variants also fit comfortably on 8–16 GB GPUs at Q4_K_M.
What quantization should I use for MiniCPM-V?
Q4_K_M is the best balance of quality and VRAM for MiniCPM-V in most cases. Choose Q8_0 for near-lossless quality if you have spare VRAM, or smaller quants (Q3/Q2) only when memory is tight.
How do I run MiniCPM-V with Ollama?
Install Ollama, then run: ollama run minicpm-v4.6. This downloads MiniCPM-V and starts a local, OpenAI-compatible endpoint — no internet connection is needed after the initial download.