News
ReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior ReleaseReleasesOVHcloud unveils Olala mini, its first in-house LLMChinaA 9B model beats a 1T model on FinQA, trained with Crusoe Managed RLReleasesThe internal metaphysics of an LLMReleasesMapping Language Model Internal Representations to Physical Laws and Scientific ReasoningResearchCan China keep rogue AI in check? Meet the start-up testing the top modelsFrontierMicrosoft aims to challenge OpenAI, Anthropic with cheaper, specialized AI modelsOpen modelsMicrosoft rolls out new first-party models and updates to agentic AI continuous improvementFrontierMicrosoft unveils AI models for coding, cybersecurity as it takes on Anthropic, OpenAIFrontierGoogle is making it easier to opt out of Gemini AI model training – here’s howFrontierOpenAI Unveils New Astra AI Model After Scrapping Prior Release
EveryLLM
Vicuna

Vicuna

An open chat model fine-tuned from LLaMA and Llama 2 by LMSYS

Vicuna is a large language model put out by LMSYS, first released on 30 March 2023. Early builds were adapted from Meta's LLaMA, while version 1.5 drew on Llama 2. The training and serving code is open source and the model has appeared in AI research work.

Share on WhatsApp

  • Released by LMSYS on 30 March 2023
  • First versions fine-tuned from Meta's LLaMA
  • v1.5, dated 1 August 2023, built on Llama 2
  • Training and serving code published as open source
  • Public online demo offered for non-commercial use

What it is

Vicuna is a large language model released by LMSYS on 30 March 2023. It is a chat-style system built by adapting existing Meta models rather than training one from scratch: the earliest versions started from LLaMA, and the v1.5 line, dated 1 August 2023, started from Llama 2. The code used to train and serve it is published openly, and researchers have made use of the model in their work. LMSYS additionally hosted a free online demo of Vicuna, restricted to non-commercial use.

How it came about

LMSYS published Vicuna at the end of March 2023. The first releases were produced by fine-tuning Meta's LLaMA, and a later v1.5 line, dated 1 August 2023, was built on Llama 2 instead. Training and serving code sits in the FastChat repository on GitHub. Alongside the model itself, the group ran a public demo for non-commercial use.

Inside the model

Vicuna is not a from-scratch architecture. LMSYS took Meta's LLaMA as the starting point for the first versions and Llama 2 for v1.5, then fine-tuned those bases to produce the chat model. Fine-tuning means further training an existing model on new data to change its behaviour. The code covering training and serving is released as open source under Apache License 2.0, while the v1.5 weights carry the Llama 2 Community License. No parameter counts, context lengths, token budgets or compute figures are given here.

What it can do

Vicuna has been used as a subject in AI research. LMSYS also offered a public online demo so people could try the model directly, for non-commercial purposes only. No benchmark scores, comparisons or stated limits are recorded here.

How to use it

The training and serving code is open source and lives in the FastChat repository on GitHub under Apache License 2.0. Weights for v1.5 are distributed under the Llama 2 Community License. LMSYS additionally ran a public online demo of the model, available for non-commercial use.

Why it matters

Vicuna has been used in AI research. It is also linked to Chatbot Arena, the LMSYS service in which members of the public compare chatbots by putting questions to two anonymous systems drawn from a pool of nine and voting on the replies, with the names revealed only after a vote. Users can replay a round or start over with different models.

Questions people ask about Vicuna

What is Vicuna?

It is a large language model from LMSYS, released on 30 March 2023. Rather than being trained from nothing, it was produced by fine-tuning Meta's LLaMA, with the v1.5 line later built on Llama 2. The code behind it is open source.

Who made it?

LMSYS, the group also behind Chatbot Arena. The model first appeared on 30 March 2023, and version 1.5 followed on 1 August 2023.

How can I get it?

Training and serving code is published openly in the FastChat repository on GitHub under Apache License 2.0. The v1.5 weights fall under the Llama 2 Community License, and LMSYS ran an online demo limited to non-commercial use.

What is it built on?

The earliest versions were fine-tuned from Meta's LLaMA. Version 1.5 instead used Llama 2 as its base. Fine-tuning means extra training of an existing model to adjust how it behaves.

Good reading and sources

Where this page comes from

The facts on this page come from the Wikipedia article "Vicuna LLM", written by its contributors and used under the CC BY-SA 4.0 licence, and from its Wikidata record (CC0). EveryLLM rewrote the text in its own words, with AI assistance, and compares every page with the article to make sure no passage is copied. It may be reused under the same CC BY-SA 4.0 licence with a link back.

It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.

The weekly digest

Every Monday: the model of the day, new pages on EveryLLM, and the conferences and anniversaries of the week ahead.