
Academic lab · Lausanne, Switzerland
EPFL LLM TeamMeditron
The EPFL LLM team built Meditron, an open suite of medical language models, and the Megatron-LLM trainer used to produce them.
Lausanne, Switzerland
- Type
- Academic lab
- Based in
- Lausanne, Switzerland
- Last activity
- 2 years ago
- Key papers
- 1
- Featured projects
- 6
About
The EPFL LLM team is a group at the Swiss federal institute in Lausanne that publishes under the epfLLM GitHub and epfl-llm Hugging Face organizations. Its Megatron-LLM repository, created in July 2023, adapts NVIDIA's Megatron-LM trainer for distributed training of open models.
In November 2023 the team released MEDITRON, 7B and 70B-parameter medical models built on Llama-2 and further pretrained on PubMed papers, abstracts and international medical guidelines. The paper reports that MEDITRON-70B outperforms GPT-3.5 and Med-PaLM on four benchmarks, and the team released both model weights and the corpus-building code.
Mission and ambitionsTo provide openly available medical language models and training tools so that clinical AI can be studied and improved in the open.
Known for
- 01
Meditron open medical LLMs
- 02
Megatron-LLM distributed trainer
- 03
Open weights and corpus-building code
Milestones
- 2023
GitHub organization created
The epfLLM organization was opened to host the team's open training code.
- 2023
Megatron-LLM
A distributed trainer for large language models was published on GitHub.
- 2023
MEDITRON-70B paper
Chen, Bosselut, Jaggi and colleagues released 7B and 70B medical models and the code to rebuild their pretraining corpus.
- 2024
Meditron project website
The team launched the official Meditron project site, with the repository gaining over 2,000 stars.
Recent activity
Dormant: last public activity 2 years ago (2024-05-20)
Based on public GitHub pushes and Hugging Face uploads, refreshed weekly.
Open source
Meditron is a suite of open-source medical Large Language Models (LLMs). (2,221 stars)
distributed trainer for LLMs (591 stars)
Official project website for Meditron, developed by the EPFL LLM team. (2 stars)
- epfl-llm/meditron-7bmodel
Model with 331 likes on the Hub.
Model with 274 likes on the Hub.
- epfl-llm/guidelinesdataset
Dataset with 167 likes on the Hub.
Research papers
Notable releases
Openness and licences
From the 2 most downloaded of 2 models on Hugging Face.
From the 3 most starred of 3 public GitHub repos.
From the 1 newest of 1 datasets on Hugging Face.
Licence tags as published on GitHub and Hugging Face, checked 2026-11-17. Open means mostly permissive or copyleft licences; Mixed means a blend; Restricted means mostly custom or non-commercial terms.
Weights, data and training code are public on Hugging Face and GitHub.
People
- Mary-Anne HartleyProfileSenior author of MEDITRON
- Zeming ChenProfileLead author of MEDITRON
- Alexandre SallinenProfileMEDITRON co-author
Cards with a gold ring and a Profile tag have a full page with public activity. Listed from the lab's own pages and public profiles.
Contact and community
Contact through the project website.
Sources
Compiled from the lab's public pages, Hugging Face and GitHub. Last checked 2026-11-17. Something wrong or out of date? Tell us on Discord.



