European Open Source AI Index

Database News Guides About Contribute

Vicuna

by LMSYS

About the model:

Vicuna is a chat assistant trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT.

Model type:

Text

Model performance class:

Full

Link to the model:

https://huggingface.co/lmsys/vicuna-13b-v1.5

Base models:

Vicuna-13B

End model:

Vicuna-13B-v1.5

End model license:

Non-commercial license

About the organisation:

According to its website, 'The Large Model Systems Organisation develops large models and systems that are open, accessible and scalable'

Link to the organisation:

https://lmsys.org/

Model release date:

March 2023

Availability

Base Model Data

Vicuna is fine-tuned LLaMA, and LLaMA in turn is based on 'publicly available datasets' that are not all specified or easily downloadable.

https://github.com/facebookresearch/llama/blob/main/MODEL_CARD.md#training-dataset

End User Model Data

From the documentation 'We will not release the ShareGPT dataset'. Also 'Vicuna v1.3 is fine-tuned from LLaMA with supervised instruction fine-tuning. The training data is around 140K conversations collected from ShareGPT.com.'

https://github.com/lm-sys/FastChat#fine-tuning

Base Model Weights

Unlike Vicuna 13B v0, these weights do not require applying delta

https://github.com/lm-sys/FastChat#vicuna-weights

End User Model Weights

Model weights made available on HuggingFace.

https://huggingface.co/lmsys/vicuna-13b-v1.5

Training Code

Actively maintained repository

https://github.com/lm-sys/FastChat

Documentation

Code Documentation

Code is quite well-documented and released as part of the FastChat framework.

https://github.com/lm-sys/FastChat

Hardware Architecture

Preprint

Preprint covers training of the Vicuna model.

https://arxiv.org/pdf/2306.05685.pdf

Paper

Paper published in NeurIPS.

https://proceedings.neurips.cc/paper_files/paper/2023/hash/91f18a1287b398d378ef22505bf41832-Abstract-Datasets_and_Benchmarks.html

Modelcard

Minimal model card, but many details are not provided or have to be pieced together from elsewhere.

https://huggingface.co/lmsys/vicuna-13b-v1.3

Datasheet

No datasheet provided.

Access

Licenses

From the documentation 'Vicuna is based on LLaMA and should be used under LLaMA's model license.'

https://github.com/lm-sys/FastChat#vicuna-weights

Is this information not up to date?

Contribute here ->

Supported by the Centre for Language Studies and the Dutch Research Council. Website design & development © 2024 by BSTN. This version of the index generated 09 April 2026, website content last updated 11 March 2026.