Skip to content
localmodel.run

Text model · MiniCPM

MI MiniCPM-V 4.6: RAM and VRAM requirements

MiniCPM-V 4.6 needs about 1.6 GB to run at Q4_K_M (the Q4_K_M GGUF file is ~0.53 GB to download; KV cache and overhead add the rest), or about 1.9 GB at Q8_0. The lightest hardware that runs it is Apple M1 (8GB).

MiniCPM family · 1.3B params · released Apr 2026.

Shopping for hardware? See what runs MiniCPM-V 4.6 →

License Apache-2.0 · Commercial OK ↓ 897.7K/mo ♥ 1.2K on HuggingFace
Q4_K_M GGUF
0.53 GB
Q8_0 GGUF
0.81 GB
Memory @ Q4 (4k)
~1.6 GB
Context
256 k

Q4_K_M and Q8_0 are GGUF download (file) sizes. Memory @ Q4 includes the KV cache and overhead, so it runs larger than the file on disk.

Will it run on your device?

MiniCPM-V 4.6 runs on 40 of 40 tracked devices at Q4_K_M.

40 run well 0 tight fit 0 too small
Fit check Q4_K_M
Yes, it runs fast
needs 1.6 GB usable 10.5 GB
See the full breakdown
$ollama run minicpm-v4.6

Memory breakdown

Weights (Q4_K_M)0.53 GB
+
KV cache (4k)0.3 GB
+
Overhead0.8 GB
=
Total1.6 GB

How context length changes it

4k context ~1.6 GB 32k context ~3.5 GB 128k context ~10.1 GB

MiniCPM-V 4.6 is small enough that the context window is the thing to size for: 128k needs ~10.1 GB versus ~1.6 GB at 4k, so pick the window you actually use.

Speed drops too: every token re-reads the KV cache, so at 128k context MiniCPM-V 4.6 generates at roughly ~6% of its short-context speed. How this is estimated.

Quantization sizes

GGUF quantson disk
Quantization Size on disk
Q2_K 0.5 GB est
Q3_K_M 0.6 GB est
Q4_K_M (default) 0.53 GB
Q5_K_M 0.9 GB est
Q6_K 1.1 GB est
Q8_0 0.81 GB
FP16 1.52 GB

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama
$ ollama run minicpm-v4.6
llama.cpp
$ llama-cli -hf openbmb/MiniCPM-V-4_6-gguf:Q4_K_M
LM Studio
$ lms get openbmb/MiniCPM-V-4_6-gguf

Which devices can run MiniCPM-V 4.6?

Apple Silicon Macs

RAM-only laptops

iPhone & iPad

Android

NVIDIA GPUs

AMD GPUs

Same job, different size

Pick by what fits your memory: step down to free up VRAM, or step up if you have headroom.

Next size down LFM2.5 1.2B Thinking 1.17B Next size up Qwen2.5 1.5B 1.54B

Similar models

FAQ

How much VRAM or RAM does MiniCPM-V 4.6 need?

At Q4_K_M, MiniCPM-V 4.6 needs about 1.6 GB (weights ~0.53 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~1.9 GB.

What is the Q4_K_M GGUF file size of MiniCPM-V 4.6?

The Q4_K_M GGUF file is about 0.53 GB to download, and the Q8_0 GGUF is about 0.81 GB. That is the weights file on disk; to run it you also need room for the KV cache and overhead, so budget ~1.6 GB of memory at Q4_K_M.

Can MiniCPM-V 4.6 run on a laptop?

Yes, MiniCPM-V 4.6 fits on a 16 GB machine at Q4_K_M and runs on Apple Silicon or a 12 GB+ GPU comfortably.

Can I use MiniCPM-V 4.6 commercially?

Yes. MiniCPM-V 4.6 is licensed Apache-2.0, which permits commercial use.

Understand the numbers

Short guides to the ideas behind MiniCPM-V 4.6's memory and quant figures.

OpenBMB's edge vision-language model: SigLIP2-400M vision tower on a Qwen3.5-0.8B LLM, ~1.3B total, 256K context. Text GGUF sizes from the vendor repo; vision needs the ~1.1GB f16 mmproj on top (the 1.6GB Ollama blob bundles it). Image, multi-image and video understanding on phone-class hardware.

Sources

Last validated 2026-08-03. Memory figures are estimates. See methodology.