Daily AI news, sourcedWhere it stands
As of Oct 12, OpenBMB released model cards for MiniCPM-V 4.7 under Apache 2.0, in a 1B variant and a 35B-A3B mixture-of-experts variant for single-image, multi-image and video understanding.
- Official
OpenBMB publishes MiniCPM-V 4.7 in 1B and 35B-A3B sizes, the first MoE model in the on-device vision series
Hugging Face (OpenBMB) published this on Oct 12, 2026 (source: huggingface.co).
OpenBMB released model cards for MiniCPM-V 4.7 under Apache 2.0, in a 1B variant and a 35B-A3B mixture-of-experts variant for single-image, multi-image and video understanding. It compresses visual tokens early inside the vision encoder and supports mixed 4x/16x visual token compression. OpenBMB says the 1B model beats Qwen3.5 0.8B in Instruct mode on overall score and inference efficiency. The models run on iOS, Android and HarmonyOS, with day-0 support for Rockchip, Houmo, Intel, AMD and Arm chips and for SGLang, vLLM, llama.cpp and Ollama.
Source: Hugging Face (OpenBMB)
- Official
- An institution announces something that has already happened, through its own channel or a founder or executive account.
- Statement by …
- A named person's views, predictions, plans or comments, including remarks reported by media when the original is unavailable.
- Reported by …
- Only media or another third party is saying it. We name the outlet.
- Unconfirmed
- Circulating online (posts, comments, leaks). No company, named person or outlet has confirmed it.
Items older than 72 hours never go in Top stories and show their original date.