#MiniCPMV
MiniCPM-V 4.7 (github.com/OpenBMB/Mini...) brings a 35B-A3B multimodal MoE model to an iPhone at 13+ tokens/s. Its key trick is Lookahead Expert Scheduling: predict which experts the next layer will need and preload their weights while the current layer computes.

#MiniCPMV #MultimodalAI
GitHub - OpenBMB/MiniCPM-V: See, hear, and interact on your local devices.
See, hear, and interact on your local devices. Contribute to OpenBMB/MiniCPM-V development by creating an account on GitHub.
github.com
October 19, 2026 at 4:02 AM
MiniCPM‑V 4.5 is an 8‑billion‑parameter multimodal model that uses 46.7% of GPU memory and 8.7% of inference time of Qwen2.5‑VL 7B, beating GPT‑4o on benchmarks. Read more: https://getnews.me/minicpm-v-4-5-advances-efficient-multimodal-ai-with-8b-parameters/ #minicpmv #multimodalai
September 25, 2025 at 6:02 PM
OpenBMB 推出了 MiniCPM-V,這是一個輕量級的 MLLM,主打在行動裝置上進行高效的圖像與影片理解。目前看下來,它在裝置端的表現不錯,可以處理影像、影片及文字輸入。同系列的 MiniCPM-o 則更進一步,支援即時的串流影音互動,並能輸出文字與語音。對於需要在行動裝置上跑多模態 LLM 的應用,這是一個值得關注的選項,但實際效果和資源消耗還需要更多實測數據。

#MiniCPMV #MLLM #OnDeviceAI

https://github.com/OpenBMB/MiniCPM-V
May 22, 2026 at 6:16 AM