OpenBMBMiniCPMMiniCPMV4_7ForConditionalGeneration

MiniCPM V 4.7 35B A3B — Hardware Requirements & GPU Compatibility

VisionFunctions

MiniCPM-V 4.7 35B-A3B is OpenBMB's vision-language model and the first mixture-of-experts entry in the MiniCPM-V series, with about 35 billion total parameters and roughly 4 billion active per token. It reads single images, multiple images, and video, and is built with phones and other edge devices in mind: the card describes a scheduling technique and layer-wise quantization that let it run on an iPhone 17 Pro Max at over 13 tokens per second. On a PC, a Q4 build fits a 24 GB GPU or a 32 GB unified-memory machine, and the small active set keeps decoding fast. The context window is 262,144 tokens. It is released under the Apache 2.0 license, so commercial use is unrestricted. The vision encoder compresses visual tokens early and supports mixed 4x and 16x compression, letting you trade image detail for speed. It works with llama.cpp, Ollama, vLLM, and SGLang, and GGUF, AWQ, GPTQ, and FP8 variants are available.

35.2B params · MoE4B active262K contextApache 2.0Oct 202617.6K downloads848 likes155.1K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Runs well on a 24 GB GPU

Runs from 12 GB (IQ2_S · 16K)Best on 24 GB (Q4_K_M · ~169 tok/s · 32K)Macs from 32 GB (~27 tok/s)

Apache 2.0

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

Run MiniCPM V 4.7 35B A3B

One recommended way. Every other method, OS and quant is in the step-by-step guide.

Other ways

each opens step-by-step

Run in cloud

Fits on RTX A6000 (48 GB) (25 GB headroom) · Q4_K_M

Generation speed: ~141 tok/s; generation speed
Cost per 1M output tokens: $0.65; per 1M output tokens

Deployon RunPodBest optionfrom $0.33/hr

Compare GPUs →

Run MiniCPM V 4.7 35B A3B

22.1 GB VRAM at 16K context · fits the RTX 4090 (1.9 GB free)

Works on macOS, Windows and Linux.

Install Ollama (Linux)
$curl -fsSL https://ollama.com/install.sh | sh
Install Ollama (macOS / Windows)
Download and run the app; it installs the ollama command.
Download Ollama ↗
Run Q4_K_M
$ollama run hf.co/openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M
Raise the context window
Type it inside the chat. For the API server instead, start it with OLLAMA_CONTEXT_LENGTH=16384 ollama serve.
$/set parameter num_ctx 16384
A larger context needs more memory than the weights alone.
Use the API
Ollama serves its API on http://localhost:11434.

Works on macOS, Windows and Linux.

Install LM Studio
Desktop app for macOS (Apple silicon), Windows and Linux.
lmstudio.ai ↗
Download the model
In the app press Ctrl/Cmd + Shift + M and paste https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-gguf, then pick Q4_K_M. Or use the CLI (run LM Studio once first):
$lms get https://huggingface.co/openbmb/MiniCPM-V-4.7-35B-A3B-gguf@Q4_K_M
Open on Hugging Face ↗
Load and chat
Open the model loader, select the model and set Context Length to 16384. Leave GPU offload at maximum if it fits.
Start the local server (optional)
$lms server start

Works on macOS, Windows and Linux.

Install llama.cpp (macOS)
MacPorts, Nix and conda-forge also work.
$brew install llama.cpp
Install options ↗
Install llama.cpp (Windows)
$winget install llama.cpp
Install llama.cpp (Linux)
Download the prebuilt llama-*-bin-ubuntu-* archive for your GPU backend (CPU, Vulkan) from the releases page, unpack it and run the binaries from that folder. Homebrew (brew install llama.cpp) and conda-forge also work on Linux.
llama.cpp releases ↗
Start the server
Downloads the model on first run. The chat UI and an OpenAI-compatible API are at http://localhost:8080.
$llama-server -hf openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M -c 16384 -ngl 99 --port 8080

Works on macOS, Windows and Linux.

Option A: Docker Model Runner
Docker Desktop: Settings > AI > Enable Docker Model Runner. Docker Engine on Linux: install the docker-model-plugin package.
$docker model run hf.co/openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M
Model Runner docs ↗
The quant tag after the repo is how Hugging Face cards write it; if your Docker version rejects it, drop the tag and pick the quant by repo.
Option B: llama.cpp server image (NVIDIA)
Needs the NVIDIA Container Toolkit. Models are cached in the llama-cache volume. Open http://localhost:8080.
$docker run --gpus all -p 8080:8080 -v llama-cache:/root/.cache ghcr.io/ggml-org/llama.cpp:server-cuda -hf openbmb/MiniCPM-V-4.7-35B-A3B-gguf:Q4_K_M -c 16384 -ngl 99 --host 0.0.0.0 --port 8080
Image tags (cuda13, vulkan, rocm) ↗
Cache path inside the image is not in the docs; remove the -v flag if the volume complains.

Linux only.

Install vLLM
Linux only; on Windows run it inside WSL2. NVIDIA GPU shown; AMD ROCm, Intel and CPU builds have their own install pages.
$uv venv --python 3.12 --seed
$source .venv/bin/activate
$uv pip install vllm --torch-backend=auto
vLLM installation ↗
Serve the model
OpenAI-compatible API at http://localhost:8000/v1.
$vllm serve openbmb/MiniCPM-V-4.7-35B-A3B --max-model-len 16384
Runs the original ~77 GB weights, not the Q4_K_M GGUF.

macOS only.

MLX supports only macOS, so follow these steps for that setup.

Install mlx-lm
$pip install mlx-lm
Chat in the terminal
$mlx_lm.chat --model mlx-community/MiniCPM-V-4.7-35B-A3B-4bit
Downloads the model on first run. The MLX repo has its own quantisation; it may not match the quant selected above.
Or start a local server
Listens on http://localhost:8080.
$mlx_lm.server --model mlx-community/MiniCPM-V-4.7-35B-A3B-4bit

Works on macOS, Windows and Linux.

Install Jan
jan.ai ↗
Find the model
Open the Hub tab, paste this model ID in Search, and choose Q4_K_M if offered.
$openbmb/MiniCPM-V-4.7-35B-A3B-gguf
Use the local API (optional)
Settings > Local API Server > Start Server. It listens on http://127.0.0.1:1337.

Rent GPUs hourly to run Q4_K_M right there. Pricing comes from partner marketplaces and varies.

Available GPU rentals for Q4_K_M, prioritized by value
GPU	Speed	$/1M tokens	Rent
RTX A6000 (48 GB)CheapestCheapest25 GB free	~141 tok/s	$0.65	Deploy on RunPodfrom $0.33/hr (opens in new tab)	48 GB · 25 GB free · ~141 tok/s · $0.65 / 1M tokens
RTX PRO 4500 (32 GB)9 GB free	~156 tok/s	$0.60	Deploy on RunPodfrom $0.34/hr (opens in new tab)	32 GB · 9 GB free · ~156 tok/s · $0.60 / 1M tokens
A40 (48 GB)25 GB free	~131 tok/s	$0.74	Deploy on RunPodfrom $0.35/hr (opens in new tab)	48 GB · 25 GB free · ~131 tok/s · $0.74 / 1M tokens
RTX A6000 (48 GB)25 GB free	~141 tok/s	$0.80	Deploy on Vast.aifrom $0.40/hr (opens in new tab)	48 GB · 25 GB free · ~141 tok/s · $0.80 / 1M tokens
RTX 5000 Ada (32 GB)9 GB free	~114 tok/s	$1.19	Deploy on RunPodfrom $0.49/hr (opens in new tab)	32 GB · 9 GB free · ~114 tok/s · $1.19 / 1M tokens
RTX 5090 (32 GB)9 GB free	~233 tok/s	$0.70	Deploy on Vast.aifrom $0.59/hr (opens in new tab)	32 GB · 9 GB free · ~233 tok/s · $0.70 / 1M tokens
RTX 6000 Ada (48 GB)25 GB free	~163 tok/s	$1.05	Deploy on Vast.aifrom $0.62/hr (opens in new tab)	48 GB · 25 GB free · ~163 tok/s · $1.05 / 1M tokens
A100 40GB17 GB free	~216 tok/s	$0.90	Deploy on Vast.aifrom $0.70/hr (opens in new tab)	17 GB free · ~216 tok/s · $0.90 / 1M tokens

Compare GPUs →

How Much VRAM Does MiniCPM V 4.7 35B A3B Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	On RTX 4090 (reference)
10.1 GB	11.3 GB	15.4 GB	11.3 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Fits: 256K context · ~253 tok/s
10.9 GB	12.2 GB	16.2 GB	12.3 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Fits: 256K context · ~244 tok/s
11.4 GB	12.7 GB	16.7 GB	13.7 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Fits: 256K context · ~239 tok/s
12.3 GB	13.5 GB	17.6 GB	11.7 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Fits: 256K context · ~231 tok/s
14.0 GB	15.3 GB	19.3 GB	15.5 GB	3-bit · i-quant — Importance-weighted 3-bit	Fits: 128K context · ~216 tok/s
14.5 GB	15.7 GB	19.8 GB	12.2 GB	2-bit · small K-quant — 2-bit small K-quant	Fits: 128K context · ~212 tok/s
14.9 GB	16.2 GB	20.2 GB	14.5 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Fits: 128K context · ~209 tok/s
15.3 GB	16.6 GB	20.6 GB	17.6 GB	3-bit · i-quant small — Importance-weighted 3-bit, small	Fits: 128K context · ~206 tok/s
15.3 GB	16.6 GB	20.6 GB	12.9 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Fits: 128K context · ~206 tok/s
15.8 GB	17.1 GB	21.1 GB	15.2 GB	3-bit small — 3-bit small quantization	Fits: 128K context · ~202 tok/s
16.2 GB	17.5 GB	21.5 GB	15.4 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Fits: 128K context · ~199 tok/s
17.6 GB	18.8 GB	22.8 GB	16.8 GB	3-bit medium — 3-bit medium quantization	Fits: 128K context · ~191 tok/s
18.0 GB	19.3 GB	23.3 GB	19.7 GB	4-bit · legacy — 4-bit legacy quantization	Fits: 128K context · ~188 tok/s
18.4 GB	19.7 GB	23.7 GB	18.1 GB	3-bit large — 3-bit large quantization	Fits: 128K context · ~185 tok/s
19.3 GB	20.6 GB	24.6 GB	18.7 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Fits: 64K context · ~180 tok/s
20.2 GB	21.5 GB	25.5 GB	21.8 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Fits: 64K context · ~175 tok/s
20.2 GB	21.5 GB	25.5 GB	19.9 GB	4-bit small — 4-bit small quantization	Fits: 64K context · ~175 tok/s
20.2 GB	21.5 GB	25.5 GB	19.8 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Fits: 64K context · ~175 tok/s
21.5 GB	22.8 GB	26.8 GB	21.2 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Fits: 32K context · ~169 tok/s
22.4 GB	23.7 GB	27.7 GB	24 GB	5-bit · legacy — 5-bit legacy quantization	Fits: 32K context · ~164 tok/s
24.6 GB	25.9 GB	29.9 GB	26.1 GB	5-bit · legacy+ — 5-bit legacy quantization with offset	Does not fit: needs 25.2 GB
24.6 GB	25.9 GB	29.9 GB	24 GB	5-bit small — 5-bit small quantization	Does not fit: needs 25.2 GB
25.5 GB	26.7 GB	30.8 GB	24.7 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Does not fit: needs 26.1 GB
29.4 GB	30.7 GB	34.7 GB	28.5 GB	6-bit, very good — 6-bit quantization, very good quality	Does not fit: needs 30 GB
35.6 GB	36.9 GB	40.9 GB	36.9 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Does not fit: needs 36.2 GB
70.8 GB	72.1 GB	76.1 GB	~70.4 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Does not fit: needs 71.4 GB

Selected Q4_K_M

4-bit medium quantization

22.1GBat 16K context

Weights: 21.1 GB
KV cache: 0.7 GB
Overhead: 0.3 GB

Fits RTX 4090 (reference) · 1.9 GB free · ~169 tok/s

≥50 tok/s needs ≥211 GB/s · RTX 5090 ~233, RTX 3090 Ti ~169

Or rent RTX A6000 from $0.33/hr.

Which Hardware Can Run MiniCPM V 4.7 35B A3B?

Q4_K_M · 21.5 GB

MiniCPM V 4.7 35B A3B (Q4_K_M) requires 21.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 28+ GB is recommended. Using the full 262K context window can add up to 10.7 GB, bringing total usage to 32.2 GB. 7 GPUs can run it, including NVIDIA GeForce RTX 5090. 103 devices with unified memory can run MiniCPM V 4.7 35B A3B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M1 Max (32 GB).

By VRAM tier · typical GPUs and Macs at Q4_K_M

Typical hardwareQuantVRAMtok/sContext that fits

8 GB · RTX 4060Doesn't fit — smallest quant needs 10.7 GB8 GB tier →

12 GB · RTX 3060 12GBIQ2_S · 12 GB · ~129 tok/s · 16K12~12916K12 GB tier →

16 GB · RTX 5060 Ti 16GBIQ3_S · 15.9 GB · ~122 tok/s · 16K15.9~12216K16 GB tier →

24 GB · RTX 4090referenceQ4_K_M · 22.1 GB · ~169 tok/s · 32K22.1~16932K24 GB tier →

Mac 32 GB · Mac Mini M4Q4_K_M · 22.1 GB · ~27 tok/s · 128K22.1~27128KMacs →

Mac 64 GB · Mac Studio M4 MaxQ4_K_M · 22.1 GB · ~69 tok/s · 256K22.1~69256KMacs →

Alternatives to MiniCPM V 4.7 35B A3B

MiniCPM family · 16 models →

Similar models

6tok/s on RTX 4090

Model	Score	VRAM	tok/s	Why it is similar
Qwen3.6 35B A3BAlibaba	107.1	22 GB	~151	same size
Llm Jp 4.1 33B Thinkingllm-jp	—	22.5 GB	~35	same size
Agnes 3.0 FlashAgnes-AI	—	22.1 GB	~35	same size
Ornith 1.5 35B A3Bornith-ai	—	22 GB	~151	same size
Agents A1InternScience	—	21.5 GB	~171	same size
Intern S2 PreviewInternLM	—	73.2 GB	—needs > 24 GB	same size

Specifications

Publisher: OpenBMB
Family: MiniCPM
Parameters: 35.2B
Active parameters: 4B per token
Architecture: MiniCPMV4_7ForConditionalGeneration
Context Length: 262,144 tokens
Vocabulary Size: 248,144
Release Date: 2026-10-07
License: Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does MiniCPM V 4.7 35B A3B need?

MiniCPM V 4.7 35B A3B requires 21.5 GB of VRAM at Q4_K_M, or 70.8 GB at BF16. Full 262K context adds up to 10.7 GB (32.2 GB total).

VRAM = Weights + KV Cache + Overhead

Weights = 35.2B × 4.8 bits ÷ 8 = 21.1 GB

KV Cache + Overhead ≈ 0.4 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

KV Cache + Overhead ≈ 11.1 GB (at full 262K context)

VRAM usage by quantization

Q4_K_M

21.5 GB

Q4_K_M + full context

32.2 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 4090 run MiniCPM V 4.7 35B A3B?

Yes, at Q5_0 (22.4 GB) or lower. Higher quantizations like Q5_1 (24.6 GB) exceed the NVIDIA GeForce RTX 4090's 24 GB.

What's the best quantization for MiniCPM V 4.7 35B A3B?

For MiniCPM V 4.7 35B A3B, Q4_K_M (21.5 GB) offers the best balance of quality and VRAM usage. Q5_0 (22.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 10.1 GB.

VRAM requirement by quantization

IQ2_XXS

10.1 GB~53%

IQ3_XS

14.9 GB~73%

Q3_K_L

18.4 GB~86%

Q4_K_M ★

21.5 GB~89%

Q5_0

22.4 GB~90%

BF16

70.8 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run MiniCPM V 4.7 35B A3B on a Mac?

Yes, but only at lower quantizations. The smallest Mac that can run MiniCPM V 4.7 35B A3B is Mac Mini M1 (16 GB) at IQ2_XXS; 135 of the 156 Macs we list can run it at some quantization. For Q4_K_M (21.5 GB) you need a Mac with more unified memory.

Can I run MiniCPM V 4.7 35B A3B locally?

Yes — MiniCPM V 4.7 35B A3B can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

How fast is MiniCPM V 4.7 35B A3B?

At Q4_K_M, MiniCPM V 4.7 35B A3B can reach ~118 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~169 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.

tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)

Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.

Example: NVIDIA B200 → 2.4 GB active ÷ (8000 × 0.65) = 0.471 ms, plus 40 layers × 0.055 ms = 2.200 ms, so 1000 ÷ 2.671 ms = ~374 tok/s

Estimated speed at Q4_K_M (21.5 GB)

NVIDIA B200

~374 tok/s

NVIDIA GeForce RTX 4090

~169 tok/s

NVIDIA B300

~374 tok/s

NVIDIA GH200 Grace Hopper Superchip

~337 tok/s

Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.

Learn more about tok/s estimation →

What's the download size of MiniCPM V 4.7 35B A3B?

At Q4_K_M, the download is about 21.13 GB. The full-precision BF16 version is 70.43 GB. The smallest option (IQ2_XXS) is 9.68 GB.

Which GPUs can run MiniCPM V 4.7 35B A3B?

7 consumer GPUs can run MiniCPM V 4.7 35B A3B at Q4_K_M. Top options include AMD Radeon RX 7900 XTX, NVIDIA GeForce RTX 3090.

Which devices can run MiniCPM V 4.7 35B A3B?

103 devices with unified memory can run MiniCPM V 4.7 35B A3B at Q4_K_M (21.5 GB), including AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10, Asus ROG Flow Z13 (2025, Ryzen AI Max+ 395, 128 GB), Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.