← sources

A 35B multimodal model running on a phone with ~3GB memory? MiniCPM-V 4.7 demo is pretty wild

source: reddit (reddit:LocalLLaMA)published 2026-10-14 15:53ingested 2026-10-15 02:42

retained text

Just saw this demo from OpenBMB for **MiniCPM-V 4.7**, and the on-device capability is pretty impressive.

A **35B-A3B MoE vision-language model** running locally on mobile devices is already interesting, but the use cases are what caught my attention:

* 📷 Reading real-world signs
* 📄 Understanding complex documents
* 🧠 Reasoning over visual information
* 🗺️ Planning with real-world constraints

The key piece seems to be **LES (Lookahead Expert Scheduling)** — instead of just making the model smaller, they optimize MoE inference by predicting and preloading experts to reduce bottlenecks.

They also released a smaller **MiniCPM-V 4.7 1B** model for more resource-constrained devices.

Another interesting addition is **Self-Replay Fine-Tuning (SRFT)**, which aims to let developers customize the model for specific tasks while keeping general capabilities.

The direction is interesting: instead of sending every image/question to the cloud, we're getting closer to having capable multimodal assistants running directly on phones.

Links:

GitHub: [https://github.com/OpenBMB/MiniCPM-V](https://github.com/OpenBMB/MiniCPM-V)  
Hugging Face: [https://huggingface.co/collections/openbmb](https://huggingface.co/collections/openbmb)  
SRFT details: [https://openbmb.cn/en/blog/self-replay-fine-tuning](https://openbmb.cn/en/blog/self-replay-fine-tuning)

document fetched 2026-10-15 00:45 · canonical https://reddit.com/r/LocalLLaMA/comments/1x5uwnd/a_35b_multimodal_model_running_on_a_phone_with