Openbmb releases MiniCPM-V 4.7-35B-A3B, first open-source MoE model for on-device use
openbmb launched MiniCPM-V 4.7-35B-A3B, the first open-source mixture-of-experts (MoE) model optimized for on-device inference on phones and tablets. The model runs smoothly on an iPhone 17 Pro Max with 13+ tokens per second and 3GB of memory, thanks to techniques like Lookahead Expert Scheduling (LES), fine-grained quantization, and DSpark speculative sampling. It supports single-image,…
Key points
- First open-source MoE model for on-device use, running at 13+ tokens/s on iPhone 17 Pro Max with 3GB memory
- Outperforms Qwen3.6 35B-A3B in Instruct mode with faster inference and higher scores, per openbmb’s benchmarks
- Supports iOS, Android, HarmonyOS, and chips from Rockchip, Houmo, Intel, AMD, and Arm with day-0 compatibility
The model supports iOS, Android, and HarmonyOS, with day-0 compatibility for chips from Rockchip, Houmo, Intel, AMD, and Arm. It also integrates with frameworks like SGLang, vLLM, llama.cpp, and Ollama, offering GGUF, BNB, AWQ, and GPTQ quantization variants, including FP8 support. openbmb provides Self-Replay Fine-Tuning (SRFT) for low-cost customization. The model has seen 630 downloads last month on Hugging Face.
openbmb/MiniCPM-V-4.7-35B-A3B
huggingface.co · 12 October 2026
Loading the full article…
This text was published by huggingface.co. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
Coverage and discussion
2sources- Reddit discussionreddit.com
- Reddit discussionreddit.com
- openbmb/MiniCPM-V-4.7-1BPrimary source · huggingface.co ·
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- Z.ai’s GLM-Legal scores 19.2% on Harvey LAB, outpacing Claude Fable 5 and Kimi K3 · 1 src
- Cloudflare releases Clef-omni, a 30B multimodal decision model for text, image, audio and video · 1 src
- vLLM Semantic Router releases Decision 3.0 multimodal models with 5 sizes under Apache-2.0 · 1 src
- Anthropic cuts Haiku 5.5 pricing to match OpenAI’s GPT-6 Luna · 1 src
- OpenAI faces concerns over loop transformers; Microsoft confirms GPT-6.1 Sol uses them · 1 src
Comments
via GitHub Discussions