Meituan-AutoML/MobileVLM

Health v2 · maintenance only

Adoption not part of the score

1370 stars · 89 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

MobileVLM is a family of compact vision language models (1.4B-3B parameters) designed to run efficiently on mobile devices, combining small language models with CLIP-style vision encoders via lightweight projectors. The repository provides model weights, training code, and evaluation tooling for the MobileVLM and MobileVLM V2 architectures.

Use cases

• run a vision language model on a smartphone
• image captioning on mobile devices
• visual question answering with a small model
• build multimodal AI apps for edge devices
• research lightweight VLM architectures
• compare small VLMs on standard benchmarks

When to choose

• you need multimodal (image+text) understanding on resource-constrained or mobile hardware
• you want open Apache-2.0 VLM weights at 1B-3B scale
• you are researching efficient VLM architectures and training recipes

When to avoid

• you need state-of-the-art accuracy regardless of model size
• you need a hosted API rather than self-hosted model code
• your project requires non-Python or non-GPU environments

Facets

library · maturity active

machine-learning deep-learning llm-inference nlp image-processing artificial-intelligence large-language-models computer-vision mobile-development python cross-platform vision-language-model multimodal mobile-ai on-device-inference llava-style mobile gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Meituan-AutoML/MobileVLM")