MiniCPM-V 4.7 35B-A3B
Open vision-language model for on-device use.
Model idopenbmb/MiniCPM-V-4.7-35B-A3B
PricePrice not published
Checked13 October 2026
No published training cutoff
At least166events since its release on 7 October 2026
2 historic, 21 major. 6 days of news.
OpenBMB doesn’t publish a training cutoff, so we count from its release date. The real gap is larger.
How to call it
3 ways to use MiniCPM-V 4.7 35B-A3B, with the model id each provider expects.
| Provider | Model id | Endpoint | Docs |
|---|---|---|---|
| Hugging Face | openbmb/MiniCPM-V-4.7-35B-A3B |
huggingface.co | |
| Hugging Face (GGUF) | openbmb/MiniCPM-V-4.7-35B-A3B-gguf |
huggingface.co | |
| Ollama | openbmb/minicpm-v4.7 |
ollama.com |
Pricing
Price not published.
No price source linked. Checked 13 October 2026.
Specs
- Input
- Text, images and video
- Output
- Text
- Context window
- Not published
- Max output
- Not published
- Open weights
- Yes
- Licence
- Apache-2.0
- Training cutoff
- Not published
- Released
- 7 October 2026
- Status
- Current
- Type
- Multimodal model
What stands out
What its maker marketed at launch, and what people found later, each with its source.
35B mixture-of-experts model in phone memory
Vendor claim: 13+ tokens/s in 3 GB of memory on an iPhone 17 Pro Max, by loading only the experts about to be used (Lookahead Expert Scheduling), layer-wise quantization and DSpark speculative sampling. Not independently tested.
Notes
35B total, about 3B active parameters; Qwen3.5/3.6 language backbone and SigLIP2 vision encoder. Image, multi-image and video input. A 1B version (openbmb/MiniCPM-V-4.7-1B-gguf and quantized builds) exists. Context length and benchmark numbers are not given in text on the model card.
vllm serve "openbmb/MiniCPM-V-4.7-35B-A3B"
# or: pip install "transformers[torch]>=5.18.0" torchvision torchcodec (AutoModelForImageTextToText)