MiniCPM V 2 6 — Hardware Requirements & GPU Compatibility
VisionMiniCPM-V 2.6 is a multimodal model from OpenBMB's series with 8.1 billion parameters, designed to handle single-image, multi-image, and video understanding tasks. Its successor's card describes MiniCPM-V 4.0 as inheriting the performance of this model across those same areas, listing it at an OpenCompass score of 65.2. The architecture pairs a SigLIP-400M vision encoder with a Qwen2-7B language model. With its 8 billion parameters and quantization, the system runs on a single consumer GPU. The weights are released under OpenBMB's MiniCPM Model License, while the code uses Apache 2.0; read the model terms before any commercial use. Published in August 2024, it has been superseded within the series by the smaller 4B MiniCPM-V 4.0, which that model's card reports as scoring higher on OpenCompass despite being roughly half the size.
Runs great on a 24 GB GPU
Runs from 8 GB (Q4_K_M · 16K)Best on 24 GB (Q4_K_M · ~123 tok/s · 32K)Macs from 16 GB (Q4_K_M · ~26 tok/s)
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run MiniCPM V 2 6
One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.
Recommended
ollama run hf.co/openbmb/MiniCPM-V-2_6-gguf:Q4_K_MQ4_K_M fits fully on the RTX 4090 · 16.1 GB free · 16K context
Download: ggml-model-Q4_K_M.gguf4.68 GBopenbmb
Ways to runOther ways
each opens step-by-stepRun in cloud
Your own private copy · Q4_K_M · 24 GB GPU
- The easiest way to try this model
- Running in a few minutes, easy setup
- Chat right here or connect any app
$5+ bonus for new users
Runs on RTX 3080 (10 GB) (2 GB headroom) · Q4_K_M
- Generation speed
- ~92 tok/s
- generation speed
- Cost per 1M output tokens
- $0.37
- per 1M output tokens
How Much VRAM Does MiniCPM V 2 6 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 3.8 GB | 9.3 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Fits: 32K context · ~173 tok/s | |
| 3.9 GB | 9.4 GB | 3-bit small — 3-bit small quantization | Fits: 32K context · ~168 tok/s | |
| 4.3 GB | 9.9 GB | 3-bit medium — 3-bit medium quantization | Fits: 32K context · ~151 tok/s | |
| 4.5 GB | 10 GB | 4-bit · legacy — 4-bit legacy quantization | Fits: 32K context · ~147 tok/s | |
| 5.3 GB | 10.9 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Fits: 32K context · ~123 tok/s | |
| 6.3 GB | 11.9 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Fits: 32K context · ~103 tok/s | |
| 7.3 GB | 12.9 GB | 6-bit, very good — 6-bit quantization, very good quality | Fits: 32K context · ~89 tok/s | |
| 8.9 GB | 14.4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Fits: 32K context · ~74 tok/s |
Selected Q4_K_M
4-bit medium quantization7.9GBat 16K context
- Weights
- 4.7 GB
- KV cache
- 3.0 GB
- Overhead
- 0.3 GB
Which Hardware Can Run MiniCPM V 2 6?
Q4_K_M · 5.3 GBMiniCPM V 2 6 (Q4_K_M) requires 5.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 7+ GB is recommended. 107 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti. 172 devices with unified memory can run MiniCPM V 2 6, including Mac Studio M5 Ultra (256 GB), Mac Studio M5 Ultra (512 GB), NVIDIA Jetson Orin Nano 8GB (Super).
12 GB
12 GB tier →fits · 4.1 GB free · up to 32K ctx
16 GPUs · 1 device family
RTX 3080 Ti~111Simulate output on RTX 3080 TiRTX 3080 12GB~111Simulate output on RTX 3080 12GBRTX 5070~82Simulate output on RTX 5070iPhone~10Simulate output on Apple iPhone 17 Pro20 GB
fits · 12.1 GB free · up to 32K ctx
1 GPU
RX 7900 XT~90Simulate output on RX 7900 XT32–40 GB
fits · 24.1 GB free · up to 32K ctx
1 GPU · 3 Mac classes · 9 device families
RTX 5090~218Simulate output on RTX 5090M5 Max*36 GB~60Simulate output on Mac Studio M5 Max (36 GB)M632 GB~22Simulate output on iMac 24" M6 (32 GB)RTX Spark N1X~37Simulate output on NVIDIA RTX Spark N1X Laptop (32 GB)Orin~25Simulate output on NVIDIA Jetson AGX Orin 32GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to MiniCPM V 2 6
MiniCPM family · 19 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| ALoDLM 8BAmazon | — | 7.6 GB | ~119 | same size |
| Llm Jp 4.1 8B Thinkingllm-jp | — | 7.6 GB | ~115 | same size |
| Ling 3.0 TinyInclusion AI | — | 5.5 GB | ~364 | same size |
| LFM2.5 8B A1BLiquid AI | — | 5.6 GB | ~335 | same size |
| Gemma 4 E4BGoogle | — | 5.6 GB | ~123 | same size |
| InternVL3 5 8BOpenGVLab | — | 8.2 GB | ~116 | same size |
Specifications
- Publisher
- OpenBMB
- Family
- MiniCPM
- Parameters
- 8.1B
- Release Date
- 2024-08-04
Frequently Asked Questions
How much VRAM does MiniCPM V 2 6 need?
MiniCPM V 2 6 requires 5.3 GB of VRAM at Q4_K_M, or 17.8 GB at BF16.
VRAM = Weights + KV Cache + Overhead
Weights = 8.1B × 4.8 bits ÷ 8 = 4.9 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.
VRAM usage by quantization
What's the best quantization for MiniCPM V 2 6?
For MiniCPM V 2 6, Q4_K_M (5.3 GB) offers the best balance of quality and VRAM usage. Q4_K_L (5.5 GB) provides better quality if you have the VRAM. The smallest option is IQ2_M at 3.0 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run MiniCPM V 2 6 on a Mac?
Yes — Mac Mini M2 Pro (16 GB) and 145 other Macs can run MiniCPM V 2 6. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 5.3 GB of usable unified memory (RAM minus macOS overhead).
Can I run MiniCPM V 2 6 locally?
Yes — MiniCPM V 2 6 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is MiniCPM V 2 6?
At Q4_K_M, MiniCPM V 2 6 can reach ~897 tok/s on AMD Instinct MI350X. On NVIDIA GeForce RTX 4090: ~123 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B200 → 8000 ÷ 5.35 × 0.65 = ~972 tok/s
Estimated speed at Q4_K_M (5.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of MiniCPM V 2 6?
At Q4_K_M, the download is about 4.86 GB. The full-precision BF16 version is 16.20 GB. The smallest option (IQ2_M) is 2.73 GB.
Which GPUs can run MiniCPM V 2 6?
107 consumer GPUs can run MiniCPM V 2 6 at Q4_K_M. Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6700M, AMD Radeon RX 6750 XT, AMD Radeon RX 5600 XT. 51 GPUs have plenty of headroom for comfortable inference.
Which devices can run MiniCPM V 2 6?
172 devices with unified memory can run MiniCPM V 2 6 at Q4_K_M (5.3 GB), including AMD Ryzen 7 8845HS (Hawk Point) Laptop, AMD Ryzen AI 7 350 (Krackan Point) Laptop, AMD Ryzen AI 9 HX 370 (Strix Point) Laptop, ASUS Ascent GX10. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.