OpenBMB releases MiniCPM-V 4.7 35B-A3B, its first open-source on-device MoE vision model
Original titleopenbmb/MiniCPM-V-4.7-35B-A3B
OpenBMB releases MiniCPM-V 4.7 35B-A3B, which it calls the first open-source MoE model to run smoothly on-device without compromising performance.
The model reaches 13+ tokens/s on iPhone 17 Pro Max and is said to beat Qwen3.6 35B-A3B on overall score in Instruct mode while using less inference time.
It is available on Hugging Face in FP8, int4, AWQ, GPTQ, and GGUF variants, plus a DSpark draft model for speculative decoding.
The source details how Lookahead Expert Scheduling, layer-wise quantization, and speculative sampling aim to run an MoE model on-device, useful context for anyone tracking on-device deployment.
Source: OpenBMB (MiniCPM) · new models on Hugging Face · huggingface.coPublished · added here