MiniCPM-V 4.7
OpenBMB image-text-to-text model on Hugging Face. Transformers weights, multimodal, MoE.
Key features
- Understands single images, multiple images, and video
- Decodes 13+ tokens/s on iPhone 17 Pro Max in 3 GB
- Supports mixed 4x and 16x visual token compression
- Deploys on iOS, Android, and HarmonyOS devices
- Works with SGLang, vLLM, llama.cpp, and Ollama
- Category
- LLM APIs
- Who it's for
- For developers running or fine-tuning multimodal models on phones and local devices.
- Tags
- modelmultimodalopen-modelshostingimage-videomobile
- Website
- huggingface.co
- Listed
- 2026-10-13
Frequently asked questions
Is MiniCPM-V 4.7 free?
Yes. MiniCPM-V 4.7 is open source and free to use.
What is MiniCPM-V 4.7 used for?
OpenBMB image-text-to-text model on Hugging Face. Transformers weights, multimodal, MoE. Typical uses: understands single images, multiple images, and video; decodes 13+ tokens/s on iPhone 17 Pro Max in 3 GB; supports mixed 4x and 16x visual token compression.
What are the best MiniCPM-V 4.7 alternatives?
Alternatives in the same category include DeepInfra, LM Studio, Mistral, Replicate, Workers AI.
Related in LLM APIs
See all →- Amazon Bedrock AWS API for foundation models, including Anthropic and Amazon models.
- Anthropic Console Console and API for Claude models.
- Azure OpenAI OpenAI models hosted on Microsoft Azure.
- Cerebras Wafer-scale inference for open models.
- Cohere API for Cohere's command and embed models.
- DeepInfra Hosted inference API for open models.
