MiniCPM V 4 needs about 3.0 GB of memory at Q4_K_M (a 2.2 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 10 quants, from 2.1 GB (Q4_0) to 3.8 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 1.1 GB.
MiniCPM V 4 – 3.6B parameter MLLM with vision and coding abilities
MiniCPM V 4 is a 3.6‑billion‑parameter model built on the llama architecture. It supports single‑image, multi‑image, and video understanding, as well as coding tasks. The model is released under the Apache‑2.0 license and is available in multiple GGUF quantizations.
Good at
- Vision tasks such as image and video analysis
- Coding assistance and code generation
- On‑device inference with low latency
Limits
- Requires 8‑k context for memory estimates
- No explicit vision‑only mode in the spec
- License is Apache‑2.0, not proprietary
Best for
- Local image captioning on mobile devices
- Generating code snippets from prompts
- Running on edge devices with limited VRAM
Who it’s for. Developers and researchers who need a compact, open‑source multimodal model for on‑device deployment. The model is suitable for mobile apps, local web demos, and lightweight inference pipelines.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_0 | 2.1 GB | 2.8 GB | Fits | Get file → | |
| Q4_K_S | 2.1 GB | 2.9 GB | Fits | Get file → | |
| Q4_K_M | 2.2 GB | 3.0 GB | Fits | Get file → | |
| Q4_1 | 2.3 GB | 3.1 GB | Fits | Get file → | |
| Q5_0 | 2.5 GB | 3.3 GB | Fits | Get file → | |
| Q5_K_S | 2.5 GB | 3.3 GB | Fits | Get file → | |
| Q5_K_M | 2.6 GB | 3.3 GB | Fits | Get file → | |
| Q5_1 | 2.7 GB | 3.5 GB | Fits | Get file → | |
| Q6_K | 3.0 GB | 3.7 GB | Fits | Get file → | |
| Q8_0 | 3.8 GB | 4.6 GB | Fits | Get file → |
For 8 GB VRAM, Q4_0 (2.08 GB) or Q4_K_S (2.09 GB) are suitable, using about 2.8–2.9 GB of VRAM at 8‑k context. If more VRAM is available, Q5_0 (2.51 GB) or Q5_K_M (2.56 GB) provide higher precision. Q6_K (2.96 GB) or Q8_0 (3.83 GB) are options for 12 GB+ VRAM setups.
What GPU do you need for MiniCPM V 4?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: the best quant that fits is Q6_K (3.0 GB file, about 3.7 GB in use).
- 6 GB VRAM: the best quant that fits is Q8_0 (3.8 GB file, about 4.6 GB in use).
Which MiniCPM V 4 quant should you choose?
Q4_K_M is the usual starting point for MiniCPM V 4: it balances file size and quality well. Here it is a 2.2 GB file.
The Q8_0 file is 3.8 GB, about 1.8 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q4_0 at 2.1 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about MiniCPM V 4
How much VRAM does MiniCPM V 4 need?
About 3.0 GB at Q4_K_M with an 8K context window. That is the 2.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run MiniCPM V 4 on an 8 GB GPU?
Yes, with Q8_0: the 3.8 GB file needs about 4.6 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of MiniCPM V 4 GGUF?
It depends on the quant: Q4_0 is 2.1 GB, Q4_K_S is 2.1 GB, Q4_K_M is 2.2 GB, Q4_1 is 2.3 GB, Q5_0 is 2.5 GB, Q5_K_S is 2.5 GB and 4 more.
What context length does MiniCPM V 4 support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.3 GB at 8K and 1.1 GB at 32K.
Where can I download MiniCPM V 4 GGUF?
The files are on Hugging Face in the openbmb/MiniCPM-V-4-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does MiniCPM V 4 use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run MiniCPM V 4 on 8 GB VRAM?
Yes, the Q4_0 or Q4_K_S quantizations use about 2.8–2.9 GB of VRAM at 8‑k context, leaving headroom for other processes.
Does MiniCPM V 4 support video input?
Yes, the model supports single‑image, multi‑image, and video understanding as part of its visual capabilities.
What license applies to MiniCPM V 4?
The model is released under the Apache‑2.0 license, allowing free use and modification.