# BAAI-DCAI/Bunny

A family of lightweight multimodal models.

Repository: https://github.com/BAAI-DCAI/Bunny
Canonical: https://ross.abutalabs.com/products/baai-dcai-bunny
Language: Python
License: Apache-2.0
License Family: permissive
Topics: mllm, chatgpt, gpt-4, multimodal-large-language-models, vlm, chinese, english
Last push: 2024-11-18T12:18:33+00:00

## Health v2 (maintenance only)
Score: 26/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 0, release rhythm 35, longevity 67
- inputs: {"age_days": 945, "days_push": 653, "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 1053, forks 75 (observed 2026-08-28T04:03:23.593624+00:00)

## What it is
Bunny is a family of lightweight multimodal vision-language models that combine plug-and-play vision encoders (EVA-CLIP, SigLIP) with language backbones like Llama-3, Phi-3-mini, and Qwen1.5. The repository provides model weights, training code, curated data, and evaluation for building efficient image-understanding chat models that support Chinese and English.

## Use cases
- run a lightweight vision-language model for image question answering
- fine-tune a small multimodal LLM on my own image-text data
- compare small VLMs against larger 7B and 13B models
- deploy a multimodal chat model locally with GGUF weights
- build an image captioning or visual chat assistant
- train a LLaVA-style model with a custom vision encoder and language backbone

## When to choose
- you need strong multimodal performance from a small (4B-8B) model
- you want to swap vision encoders and language backbones plug-and-play
- you need high-resolution image input up to 1152x1152
- you want open training data and a reproducible training tutorial
- you need bilingual (Chinese/English) visual chat

## When to avoid
- you need text-only LLM inference without image input
- you need production-grade serving infrastructure rather than research code
- you need audio, video, or speech modality support
- you need a managed API instead of self-hosted weights

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, machine-learning, deep-learning, image-processing, chatbot
- domain: large-language-models, computer-vision, artificial-intelligence
- platform: python, cross-platform
- tags: vision-language-model, multimodal, lightweight-models, llava-style, image-understanding, gguf, huggingface, chinese-english, visual-question-answering, model-training, natural-language-processing, gpu

## Member repositories
- BAAI-DCAI/Bunny (main) score 26

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:23.593624+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 2026-08-30T06:59:18.728769+00:00, confidence not recorded.
  - readme: https://github.com/BAAI-DCAI/Bunny (fetched 2026-08-28T04:03:23.593624+00:00, sha 6cb0b5f4ed02)
- Data as of 2026-08-30T08:39:29.467469+00:00.