Models / MiniCPM-V 4.7
Best home rig for MiniCPM-V 4.7
Quick answer
How much VRAM does MiniCPM-V 4.7 need?
The weights are about 21.4 GB at Q4_K_M (35B params × 4.89 bits ÷ 8, before context/KV cache). Add memory for context (KV cache) and the runtime. The smallest-memory setup with a measured result here, fully in memory, is 2x RTX 3090 (48 GB) at 198.7 tok/s.
What is the fastest measured hardware for MiniCPM-V 4.7?
2x RTX 3090: 198.7 tok/s generation (vLLM TP2, 1 stream, INT4 + DSpark), from https://github.com/tonyd2wild/MiniCPM-V-4.7-35B-A3B-2x-RTX-3090.
Last updated:
Current model35B total · 3B active1 measured result model info ↗
Will it fit my GPU? Best value GPU per euro →
Real, measured generation speeds for MiniCPM-V 4.7 on home hardware, fastest first. Different quants/engines aren't directly comparable.
| Build | Quant | Engine | tok/s | Prompt processing (tok/s) | Cost | Source |
|---|---|---|---|---|---|---|
| 2x RTX 3090: MiniCPM-V 4.7 INT4 on vLLM TP2, 198.7 tok/s prose (Tech2wild) | INT4 + DSpark | vLLM TP2, 1 stream | 198.7 ⚡ | – | github.com ↗ |
Prompt processing = prefill speed reported by the source, @ prompt length. * est. TTFT @8k is calculated from measured prefill, not measured. Why it matters.
Memory math: how much VRAM / RAM does MiniCPM-V 4.7 need?
MiniCPM-V 4.7 has 35B parameters, about 3B active per token (source: OpenBMB MiniCPM-V 4.7 MoE vision (35B-A3B); int4 pack used in Tech2wild recipe). Weight size = parameters × bits-per-weight ÷ 8, before context (KV cache) and runtime overhead.
| Format | Bits/weight | Weights | Smallest common memory size that holds the weights alone |
|---|---|---|---|
| Q8_0 | 8.50 | 37.2 GB | 48 GB |
| Q6_K | 6.56 | 28.7 GB | 32 GB |
| Q5_K_M | 5.70 | 25.0 GB | 32 GB |
| Q4_K_M | 4.89 | 21.4 GB | 24 GB |
| Q3_K_M | 4.00 | 17.5 GB | 24 GB |
| Q2_K | 3.16 | 13.8 GB | 16 GB |
Bits-per-weight for GGUF quants are llama.cpp's own figures (llama.cpp quantize README). "As measured" rows apply the source's stated bits-per-weight to the parameter count; real files differ slightly because some layers stay at higher precision. Try other combinations in the VRAM calculator.
Why it can still run on smaller GPUs: it is a mixture-of-experts model, so only about 3B of its 35B parameters are used for each generated token. Runtimes such as llama.cpp (--n-cpu-moe), ExLlamaV3 or Project Maya can keep most expert weights in system RAM and only the shared layers and hot experts in VRAM. Then you need VRAM + system RAM ≥ weights, and speed is limited mainly by RAM bandwidth.
FAQ
How much VRAM does MiniCPM-V 4.7 need?
The weights are about 21.4 GB at Q4_K_M (35B params × 4.89 bits ÷ 8, before context/KV cache). Add memory for context (KV cache) and the runtime. The smallest-memory setup with a measured result here, fully in memory, is 2x RTX 3090 (48 GB) at 198.7 tok/s.
What is the fastest measured hardware for MiniCPM-V 4.7?
2x RTX 3090: 198.7 tok/s generation (vLLM TP2, 1 stream, INT4 + DSpark), from https://github.com/tonyd2wild/MiniCPM-V-4.7-35B-A3B-2x-RTX-3090.
Which GPUs have measured results for MiniCPM-V 4.7?
2x RTX 3090.
Can it run on… 2x RTX 3090
Related: what GPU do you need for 2026 models · all builds