Skip to content

MiniCPM V 4.7 35B A3B on RTX 5090: VRAM, speed and fit

MiniCPM V 4.7 35B A3B at 6-bit runs on the RTX 5090 with 120K of context at about 227 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on RTX 5090 · 6-bit (Q6_K) · Ollama / LM Studio · Chat

GPU memory
30.3 GB
of 32.0 GB
Context
120K
auto
Speed
227 tok/s
156 when full
First reply
0.1 s
2K prompt

Where the memory goes

Model
26.9 GB
Context
2.4 GB
Overhead
1.0 GB
OS
0.0 GB
Free
1.7 GB
Memory
30.3 GB of 32.0 GB
Context
120K tokens · auto
Speed
227 tok/s · 156 when full
First reply
0.1 s for a 2K prompt
Running cost
$9.30/mo at home · $58/mo on a cloud 2× RTX 2000 Ada

Validation: Validated

What MiniCPM V 4.7 35B A3B on RTX 5090 was checked against
CheckTargetResultStatus
Fits in GPU memory32.0 GB usable30.3 GB neededPass
Headroom5% free1.7 GB free (5%)Pass
Context for chat16K or more120KPass
Writing speed15 tok/s or more227 tok/sPass
First replyunder 3 s, 2K prompt0.1 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
MiniCPM V 4.7 35B A3B
Parameters
35.2B, 3B active
Precision
6-bit (Q6_K)
Context
120K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
RTX 5090
GPU can use
32.0 GB of 32.0 GB

Running cost

At home
$9.30/mo
425 W at the wall
Renting instead
$58/mo
2× RTX 2000 Ada at $0.48/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.

How to run it

llama.cpp
llama-server -hf openbmb/MiniCPM-V-4.7-35B-A3B-gguf
vLLM
vllm serve openbmb/MiniCPM-V-4.7-35B-A3B
MLX (Mac)
mlx_lm.chat --model mlx-community/MiniCPM-V-4.7-35B-A3B-mxfp4

MiniCPM V 4.7 35B A3B on similar machines

All machines: MiniCPM V 4.7 35B A3B VRAM requirements. Everything that runs here: what can the RTX 5090 run?

Other models that run on RTX 5090

Questions

Can MiniCPM V 4.7 35B A3B run on RTX 5090?

MiniCPM V 4.7 35B A3B at 6-bit runs on the RTX 5090 with 120K of context at about 227 tokens a second.

Report VR-1TI757C, calculated Nov 22, 2026 from published specs, not benchmarked.