# NExT-GPT/NExT-GPT

Code and models for ICML 2024 paper, NExT-GPT: Any-to-Any Multimodal Large Language Model

Repository: https://github.com/NExT-GPT/NExT-GPT
Canonical: https://ross.abutalabs.com/products/next-gpt
Homepage: https://next-gpt.github.io/
Language: Python
License: BSD-3-Clause
License Family: permissive
Topics: chatgpt, foundation-models, gpt-4, instruction-tuning, large-language-models, llm, multi-modal-chatgpt, multimodal, visual-language-learning, mllm
Last push: 2025-05-13T09:57:47+00:00

## Health v2 (maintenance only)
Score: 37/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 21, release rhythm 35, longevity 78
- inputs: {"age_days": 1099, "days_push": 477, "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 3638, forks 360 (observed 2026-08-28T04:08:12.761184+00:00)

## What it is
NExT-GPT is an end-to-end any-to-any multimodal large language model that accepts and generates arbitrary combinations of text, image, video, and audio. The repository provides training code, instruction-tuning datasets (MosIT), and model checkpoints built on a pretrained LLM with multimodal encoders and diffusion decoders.

## Use cases
- build a chatbot that understands and generates text, images, video, and audio
- run an any-to-any multimodal LLM inference
- fine-tune a multimodal LLM with modality-switching instruction tuning
- research cross-modal understanding and generation in LLMs
- download pretrained multimodal LLM checkpoints
- extend an LLM to new input and output modalities

## When to choose
- you need a single model handling any-to-any combinations of text, image, video, and audio
- you want a research-grade multimodal LLM with released weights and training data
- you are studying modality-switching instruction tuning or low-cost adaptor tuning

## When to avoid
- you need a production-ready, commercially supported multimodal assistant
- you only need single-modality text generation from a standard LLM
- you lack GPU resources, since the 7B model plus diffusion decoders are heavy to run

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, llm-inference, llm-training, image-processing, audio-processing, video-processing, chatbot, agent-framework
- domain: large-language-models, artificial-intelligence, deep-learning, computer-vision
- platform: python
- tags: multimodal, any-to-any, instruction-tuning, diffusion-decoders, research-code, icml-2024, natural-language-processing, audio, video, ai-agents, gpu, linux

## Member repositories
- NExT-GPT/NExT-GPT (main) score 37

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:12.761184+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 2026-08-29T18:31:49.343537+00:00, confidence not recorded.
  - readme: https://github.com/NExT-GPT/NExT-GPT (fetched 2026-08-28T04:08:12.761184+00:00, sha 2cbeac44130f)
  - homepage: https://next-gpt.github.io/ (fetched 2026-08-29T09:26:16.538501+00:00, sha 2285d3eeacc9)
- Data as of 2026-08-30T08:39:29.467469+00:00.