Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

MiniCPM-V 4.7 (35B-A3B and 1B) GPU requirements: VRAM and cheapest GPU

MiniCPM-V 4.7 (35B-A3B and 1B) has 35.2B parameters, published in BF16. It is a mixture-of-experts model: 8 of its 256 experts work on each token, about 3B parameters, so it generates at roughly the speed of a model that size. Every expert still has to sit in GPU memory, which is why the memory figures below are for the full model. It is released under apache-2.0, which allows commercial use.

Parameters
35.2B
Smallest footprint (INT4)
about 22 GB
Cheapest GPU per hour (FP16)
$0.50
Type
Vision
  • Parameters 35.2B
  • Active per token 3B
  • Type Vision
  • Task image-text-to-text
  • Max context 256K
  • Licence apache-2.0
  • Downloads 876.3K
  • Likes 1.1K

Calculating for MiniCPM-V 4.7 (35B-A3B and 1B)

Precision
Workload
More settings
Serving engine
Requests at once

MiniCPM-V 4.7 (35B-A3B and 1B) needs about 79 GB of GPU memory

Inference (vLLM), FP16 / BF16, 8K tokens context, one request at a time.

Where the memory goes

  • Weights 70 GB
  • KV cache 0.17 GB
  • Runtime overhead 8.1 GB
  • Total 79 GB

GPU counts keep 10% of each card free for spikes. Plan extra if you expect long prompts or many users at once.

GPUs that fit, cheapest firstEvery GPU type we track that can hold this setup on one node, ranked by total hourly cost
RankConfigurationPer GPU / hrGPUs neededTotal / hrTotal / month (24/7)Provider
11 × RTX PRO 6000 96GBCheapestBlackwell · GDDR7$0.501$0.50$365RunPod (opens in a new tab)1x RTX PRO 6000 MaxQ
24 × RTX 3090 24GBAmpere · GDDR6X$0.244$0.96$702Vast.ai (opens in a new tab)4x RTX 3090 (median of 9 offers)
32 × A40 48GBAmpere · GDDR6$0.592$1.18$861RunPod (opens in a new tab)2x A40
42 × RTX A6000 48GBAmpere · GDDR6$0.692$1.38$1,007RunPod (opens in a new tab)2x RTX A6000
52 × L40S 48GBAda Lovelace · GDDR6$0.802$1.60$1,169Vast.ai (opens in a new tab)2x L40S (median of 6 offers)
61 × GH200 96GBHopper · HBM3$1.991$1.99$1,453Vultr (opens in a new tab)vbm-72c-480gb-gh200-gpu (bare metal)
72 × RTX 6000 Ada 48GBAda Lovelace · GDDR6$1.092$2.18$1,591RunPod (opens in a new tab)2x RTX 6000 Ada
84 × RTX 4090 24GBAda Lovelace · GDDR6X$0.584$2.32$1,696Vast.ai (opens in a new tab)4x RTX 4090 (median of 12 offers)
91 × MI300X 192GBCDNA 3 · HBM3$2.391$2.39$1,745RunPod (opens in a new tab)1x MI300X
102 × A100 80GBAmpere · HBM2e$1.202$2.40$1,754Vast.ai (opens in a new tab)2x A100 SXM4 (median of 12 offers)
113 × A100 40GBAmpere · HBM2$0.623sold as 4 GPUs; you need 3$2.46$1,799Vast.ai (opens in a new tab)4x A100 SXM4 (median of 5 offers)
124 × L4 24GBAda Lovelace · GDDR6$0.694$2.76$2,015RunPod (opens in a new tab)4x L4
131 × H200 141GBHopper · HBM3e$2.901$2.90$2,117Lium (opens in a new tab)1x H200 NVL
141 × H100 NVL 94GBHopper · HBM3$3.191$3.19$2,329RunPod (opens in a new tab)1x H100 NVL
153 × RTX 5090 32GBBlackwell · GDDR7$0.803sold as 4 GPUs; you need 3$3.20$2,338Vast.ai (opens in a new tab)4x RTX 5090 (median of 35 offers)
162 × H100 80GBHopper · HBM3$2.142$4.27$3,119Vast.ai (opens in a new tab)2x H100 SXM (median of 5 offers)
171 × B200 180GBBlackwell · HBM3e$5.601$5.60$4,088Lium (opens in a new tab)1x B200
184 × A10 24GBAmpere · GDDR6$1.424$5.67$4,141AWS (opens in a new tab)g5.12xlarge
196 × V100 16GBVolta · HBM2$0.796sold as 8 GPUs; you need 6$6.32$4,614Lambda (opens in a new tab)8x Tesla V100
206 × T4 16GBTuring · GDDR6$0.986sold as 8 GPUs; you need 6$7.82$5,712AWS (opens in a new tab)g4dn.metal
211 × B300 288GBBlackwell Ultra · HBM3e$8.001$8.00$5,840Lium (opens in a new tab)1x B300 SXM6
221 × GB200 186GBBlackwell · HBM3e$10.501sold as 4 GPUs; you need 1$42.00$30,660CoreWeave (opens in a new tab)GB200 NVL72
Cost per cloudCheapest offer that fits, per provider. Approximate, prices collected Nov 16, 2026.
ProviderInstancePer GPU / hrGPUs neededTotal / hrTotal / month (24/7)
RunPod (opens in a new tab)Cheapest1x RTX PRO 6000 MaxQus$0.501 × RTX PRO 6000 96GB about $0.50about $365
Vast.ai (opens in a new tab)4x RTX 3090 (median of 9 offers)worldwide marketplace$0.244 × RTX 3090 24GB about $0.96about $702
Lium (opens in a new tab)1x RTX PRO 6000 Blackwell Server Editionworldwide marketplace$1.191 × RTX PRO 6000 96GB about $1.19about $869
Nebius (opens in a new tab)1x RTX PRO 6000 96GBeu/us$1.801 × RTX PRO 6000 96GB about $1.80about $1,314
Vultr (opens in a new tab)vbm-72c-480gb-gh200-gpu (bare metal)us$1.991 × GH200 96GB about $1.99about $1,453
Lambda (opens in a new tab)2x A6000us$1.092 × RTX A6000 48GB about $2.18about $1,591
AWS (opens in a new tab)g7e.2xlargeus-east-1$3.361 × RTX PRO 6000 96GB about $3.36about $2,455
Google Cloud (opens in a new tab)g2-standard-48us-central1$1.004 × L4 24GB about $4.00about $2,921
CoreWeave (opens in a new tab)GH200us$6.501 × GH200 96GB about $6.50about $4,745
Azure (opens in a new tab)NC40ads_H100_v5eastus$6.981 × H100 NVL 94GB about $6.98about $5,095

Prices collected Nov 16, 2026. Every cost is approximate: on-demand, US dollars, mostly US regions (marketplaces such as Vast.ai and Lium are worldwide).

Memory by precision and workload

With a 8K-token context and one request at a time. LoRA and full fine-tuning start from 16-bit weights and QLoRA from 4-bit, so the other cells are blank.
PrecisionInferenceLoRA fine-tuneQLoRA fine-tuneFull fine-tune
FP16 / BF1679 GB93 GB–623 GB
INT842 GB–––
INT422 GB–36 GB–

Cheapest GPU by precision

Inference with an 8K context and one request at a time.

FP16 / BF16

about 79 GB · 8K tokens

The weights as published. Nothing lost, but the largest footprint.

1 × RTX PRO 6000 96GB

RunPod (opens in a new tab) · 1x RTX PRO 6000 MaxQ

Per GPU / hr
$0.50
GPUs needed
1
Total / hr
about $0.50
per month, 24/7, approx.
$365

INT8

about 42 GB · 8K tokens

Weights at half the size. Most chat and RAG work shows no visible difference.

1 × RTX A6000 48GB

Lium (opens in a new tab) · 1x RTX A6000

Per GPU / hr
$0.42
GPUs needed
1
Total / hr
about $0.42
per month, 24/7, approx.
$307

INT4

about 22 GB · 8K tokens

Weights at a quarter of the size. Small drops show up first in maths and code.

1 × RTX 3090 24GB

Lium (opens in a new tab) · 1x RTX 3090 Ti

Per GPU / hr
$0.16
GPUs needed
1
Total / hr
about $0.16
per month, 24/7, approx.
$117

Inference vs fine-tuning

  • Inference (FP16)

    1×

    of the memory inference needs

    79 GB

    1 × RTX PRO 6000 96GB · about $0.50

  • LoRA fine-tune

    1.2×

    of the memory inference needs

    93 GB

    3 × A40 48GB · about $2.36

  • QLoRA fine-tune

    0.5×

    of the memory inference needs

    36 GB

    1 × RTX A6000 48GB · about $0.42

  • Full fine-tune

    7.9×

    of the memory inference needs

    623 GB

    4 × MI300X 192GB · about $9.56

Serving it at FP16 fits on 1 × RTX PRO 6000 96GB (RunPod). LoRA needs 3 × A40 48GB (RunPod), QLoRA 1 × RTX A6000 48GB (Lium) and a full fine-tune 4 × MI300X 192GB (RunPod). Training keeps activations and optimiser state in memory as well as the weights, which is why it so often needs more GPUs than serving the same model.

Will it run on a Mac or a single RTX 4090?

RTX 4090 (24 GB)

Yes, at INT4 with an 8K context: about 22 GB fits in the card’s 24 GB.

Mac with Apple silicon

Yes, with enough memory. At INT4 it needs about 22 GB, so a Mac with 32 GB or more of unified memory (macOS lets the GPU use about three quarters of it). Expect slower generation than on a data-centre GPU.

Run it yourself or pay per token?

Run it yourself

The cheapest setup we found at FP16 is 1 × RTX PRO 6000 96GB on RunPod, about $365 a month if it runs all the time. Worth it when traffic is steady or the data must stay on your own machines.

Use a hosted API

Our review of MiniCPM-V 4.7 (35B-A3B and 1B) covers where to use it through an API and what that costs. Paying per token usually works out cheaper for light or uneven traffic.

Read our review

Similar models

Open models of a similar size.

Calculate another model

MiniCPM-V 4.7 (35B-A3B and 1B): common questions

How much VRAM does MiniCPM-V 4.7 (35B-A3B and 1B) need?

About 79 GB at FP16, 42 GB at INT8 and 22 GB at INT4, with an 8K context and one request at a time. Longer contexts and more parallel requests add KV cache on top.

What is the cheapest GPU to run MiniCPM-V 4.7 (35B-A3B and 1B)?

1 × RTX PRO 6000 96GB (about $0.50 per GPU / hour) on RunPod, at about $0.50 an hour for FP16 / BF16 inference with an 8K context. Prices collected Nov 16, 2026; they change often.

Can I fine-tune MiniCPM-V 4.7 (35B-A3B and 1B) on a single GPU?

Yes, with QLoRA: it needs about 36 GB, which fits one 80 GB GPU. LoRA needs about 93 GB and a full fine-tune about 623 GB.

How are these numbers calculated?

Weights are parameters times bytes per parameter (2 at FP16, 1 at INT8, 0.5 at INT4). The KV cache comes from the model’s layers, KV heads, head size and context length. We add about 10% plus 1 GB for the runtime and fill each GPU to 90%. Prices are on-demand list prices in US regions, collected from each provider’s public pricing.

Can MiniCPM-V 4.7 (35B-A3B and 1B) run on a consumer GPU like the RTX 4090?

Yes, at INT4 with an 8K context: about 22 GB fits in the card’s 24 GB.

Does MiniCPM-V 4.7 (35B-A3B and 1B) fit on a single H100 80GB?

Yes, at INT8, INT4 with an 8K context.

How many GPUs do I need for MiniCPM-V 4.7 (35B-A3B and 1B)?

At FP16 with an 8K context: 2 × 80 GB, 2 × 48 GB or 4 × 24 GB GPUs. At INT4 you need about a quarter of the memory for the weights.

How much VRAM does MiniCPM-V 4.7 (35B-A3B and 1B) need at INT8?

About 42 GB: 37 GB for the weights, 0.17 GB for an 8K KV cache and 4.7 GB of runtime overhead.

How much does it cost to run MiniCPM-V 4.7 (35B-A3B and 1B) 24/7 for a month?

About $365 on the cheapest setup we found (1 × RTX PRO 6000 96GB (about $0.50 per GPU / hour), $0.50 an hour for 730 hours). The cheapest on AWS, Google Cloud or Azure is about $2,455 a month.

FP16 vs INT8 vs INT4: which should I use?

FP16 keeps full quality and needs the most memory (about 79 GB here). INT8 halves the weights with little loss for most chat work. INT4 (AWQ, GPTQ or GGUF Q4) cuts them to a quarter, about 22 GB in total, with a small drop in quality that shows most in maths and code.

Can MiniCPM-V 4.7 (35B-A3B and 1B) run on a Mac?

Yes, with enough memory. At INT4 it needs about 22 GB, so a Mac with 32 GB or more of unified memory (macOS lets the GPU use about three quarters of it). Expect slower generation than on a data-centre GPU.

vLLM or llama.cpp?

vLLM is built for serving many requests at once on NVIDIA and AMD GPUs and reserves most of the GPU memory for KV cache. llama.cpp runs GGUF files on almost anything, including Macs and CPUs, and suits one user or a small team. Memory needs are similar for the same precision.

Prices collected Nov 16, 2026. Every cost is approximate: on-demand, US dollars, mostly US regions (marketplaces such as Vast.ai and Lium are worldwide).

How we calculate

  • Weights. Parameters times bytes per parameter: 2 at FP16, 1 at INT8, 0.5 at INT4, plus about 5% for the quantisation scales at INT8 and INT4.
  • KV cache. 2 × layers × KV heads × head size × context × requests × 2 bytes, from the model’s config.json. Where a setting is missing we estimate it and say so.
  • Overhead. About 10% plus 1 GB for the runtime, in line with what vLLM reserves.
  • Fine-tuning. LoRA keeps 16-bit weights and adds a small adapter, its optimiser and activations. QLoRA does the same over 4-bit weights. A full fine-tune needs about 16 bytes per parameter with mixed-precision Adam (weights, gradients and optimiser states), plus activations.
  • GPUs. We fill each GPU to 90% of its memory and stop at 8 per node. Costs are on-demand list prices in US regions; marketplace clouds show the median offer.