# Vision-CAIR/MiniGPT-4

Open-sourced codes for MiniGPT-4 and MiniGPT-v2 (https://minigpt-4.github.io, https://minigpt-v2.github.io/)

Repository: https://github.com/Vision-CAIR/MiniGPT-4
Canonical: https://ross.abutalabs.com/products/minigpt-4
Homepage: https://minigpt-4.github.io
Language: Python
License: BSD-3-Clause
License Family: permissive
Last push: 2024-09-02T09:07:21+00:00

## Health v2 (maintenance only)
Score: 30/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 0, release rhythm 35, longevity 88
- inputs: {"age_days": 1236, "days_push": 730, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 25627, forks 2868 (observed 2026-08-28T04:11:38.371170+00:00)

## What it is
Official code for MiniGPT-4 and MiniGPT-v2, vision-language models that align a frozen visual encoder with a frozen LLM (Vicuna) via a single projection layer. It enables multimodal chat capabilities like detailed image description, visual question answering, and grounded image-text generation.

## Use cases
- generate detailed descriptions of images with an open-source model
- build a visual chatbot that answers questions about images
- run multimodal vision-language research experiments
- caption images using a GPT-4-like open model
- fine-tune a vision-language model on custom image-text data
- create websites or stories from handwritten drafts and images

## When to choose
- you want an open-source, computationally efficient GPT-4-style multimodal model
- you need a research baseline for vision-language multi-task learning
- you want to fine-tune only a small projection layer instead of a full LLM
- you need reproducible code from a published paper with pretrained checkpoints

## When to avoid
- you need a production-grade, actively maintained multimodal model
- you lack a GPU or cannot run large language models locally
- you want the latest multimodal capabilities rather than a 2023-era research model
- you need commercial licensing beyond BSD-3-Clause terms for model weights

## Facets
- artifact type: library
- maturity: maintenance
- function: machine-learning, deep-learning, llm-inference, llm-training, image-processing, nlp
- domain: artificial-intelligence, large-language-models, computer-vision, deep-learning, machine-learning
- platform: python
- tags: vision-language-model, multimodal, image-captioning, visual-question-answering, research-code, vicuna, gradio-demo, gpu, linux, docker

## Member repositories
- Vision-CAIR/MiniGPT-4 (main) score 30

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:38.371170+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:56:01.215564+00:00, confidence not recorded.
  - readme: https://github.com/Vision-CAIR/MiniGPT-4 (fetched 2026-08-28T04:11:38.371170+00:00, sha df2a5b07b302)
  - homepage: https://minigpt-4.github.io (fetched 2026-08-29T07:52:41.486165+00:00, sha d35c20cdc67e)
- Data as of 2026-08-30T08:39:29.467469+00:00.