Skip to main content
Back

MiniCPM-V Multimodal Series

MiniCPM-V is a series of multimodal vision-language models built for on-device deployment, supporting image and video understanding, with sizes ranging from 1B to 35B parameters.

Series Overview​

ModelParametersArchitectureVRAMHighlightsHuggingFace
MiniCPM-V 4.7 35B-A3B35.2B total · ~3.5B activeSigLIP canvas M-RoPE + Qwen3.5-MoE—First MoE in the series, runs smoothly on deviceDownload
MiniCPM-V 4.7 1B1.55BSigLIP canvas M-RoPE + Qwen3.5-0.8B—256K context, on-device friendlyDownload
MiniCPM-V 4.61.3BSigLIP2 + Qwen3.5-0.8B4 GBEarly ViT compression, on-device deploymentDownload
MiniCPM-V 4.6 Thinking1.3BSigLIP2 + Qwen3.5-0.8B4 GBDeep reasoning variantDownload
MiniCPM-V 4.58.7BSigLIP2 + Qwen3-8B~28 GBSurpasses GPT-4o-latest, 30+ languagesDownload
MiniCPM-V 4.04.1BSigLIP2 + MiniCPM4-3B—Surpasses GPT-4.1-miniDownload
ScenarioRecommended model
Mobile / ultra-lightweight deploymentMiniCPM-V 4.7 1B
High-throughput serving / peak performanceMiniCPM-V 4.7 35B-A3B
High-accuracy image/video understandingMiniCPM-V 4.5
Deep reasoning (math/code)MiniCPM-V 4.7 (enable_thinking) or MiniCPM-V 4.6 Thinking
Balanced performanceMiniCPM-V 4.0

On-device Deployment​

MiniCPM-V 4.7 and 4.6 can run fully offline on mobile devices (iOS / Android / HarmonyOS). Download: MiniCPM-V-Apps

TopicLink
Code repository and release notesOpenBMB/MiniCPM-V
MiniCPM-V 4.7 35B-A3B model weightsHugging Face / ModelScope
MiniCPM-V 4.7 1B model weightsHugging Face / ModelScope
Model documentationMiniCPM-V 4.6 / MiniCPM-V 4.5
Online demoMiniCPM-V 4.7 Demo
On-device apps and downloadsMiniCPM-V-Apps downloads
API accessAPI guide
Technical reports and architecture papersMiniCPM-V 4.5 / LLaVA-UHD v4
Multimodal technical community groupJoin the MiniCPM-V community

On this page

Developer ShowcaseProjects from community developersCommunity EventsJoin the OpenBMB open-source community