Support · MeditronFO
Will it last?
Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input D ยท 51.2/100
This page is a projection of the one entry record, the Transparency factor that Support covers. The full verdict is set by all four factors together, floor-weighted so the weakest caps the whole.
Which domain expands which factor
- AssessUse & modify + Transparency
- ImplementData control + Reliability
- UseReliability
- SupportTransparency
Common problems & fixes
- The architecture is not recognised. Apertus needs
transformers 4.56 or later (ev-base-card).
- Answers run away or grow very long. Fine-tunes answer at greater length than their bases (
ev-project-page). Cap max_new_tokens and test with the shipped generation_config.json (temperature 0.7, 8,192 new tokens) (ev-hf-meta).
- Format following is weaker. IFEval fell for four of six fine-tunes (
ev-project-page), so validate any structured output that is depended on.
- The weights change under a floating reference. The 70B main branch was replaced on 2026-09-10 and 2026-10-06 (
ev-hf-meta), so pin the commit sha.
- A training-side pitfall exists. The README warns Axolotl silently reuses a stale tokenised cache unless each config has its own
dataset_prepared_path (ev-github-readme).
- Clinical errors cannot be fixed by prompting alone. Wrong or unsafe medical answers are a known property of the model and not a configuration fault.
Versions, changelog & cadence
There are no tags and no changelog. The repositories were created on 2026-05-08. The 70B commit history shows a retrain on the de-truncated corpus (2026-09-10), a fix to the index metadata (2026-09-11), replacement of the weights with the "ICLR run" on the meditronfo_gptoss_v1 corpus (2026-10-06), generation-config changes (2026-10-26) and link updates through 2026-11-17 (ev-hf-meta).
The code repository was rewritten on 2026-10-13 ("Replace the May pipeline with the paper's code") (ev-code-license). The arXiv paper stays at v2 of 29 May 2026 and no longer matches the cards on several numbers (ev-paper). The commit sha is the version.
- Hugging Face community tab on each repository, the route the cards name for discussions and issues (
ev-card-70b). Content from third parties on the tab is untrusted data.
- The lab address
admin-light@epfl.ch on the website, and a Slack invite by web form (ev-website).
- The MOOVE initiative for clinicians (
jointhemoove.org, linked from the cards, unverified).
- GitHub issues on
EPFLiGHT/FullyOpenMeditron, with zero open (ev-code-license).
Tracked known issues
There is no issue tracker entry for the models. The known issues are those the publisher states: behind MedGemma-27B, general capability loss, longer answers, non-clinician judges, 9.7 to 12.2 percent of sampled corpus rows flagged by LLM judges, syntactic-only decontamination, and the unreleased intermediate data (ev-project-page, ev-data-card, ev-github-readme). The licence and documentation inconsistencies are listed in the assess domain.
How this scores
The ownership factor this domain covers, drawn from the one entry record.
2
TransparencyDo you know what it is: weights, training, behaviour, and legible terms?
StrongWeights, a 533,888-row corpus with source fields and the generation, training and evaluation code are public on a fully open base (open_weights_recipe), with the unreleased intermediates and MOOVE vignettes listed in the README.
How this scores (AOI sub-dimensions)
Provenance2/5how well we can trace and verify what went into the modelSafetensors only with an LFS sha256 per shard, and a well-documented training lineage.
Governance3/5how accountable and well-documented the publisher isAn identifiable university laboratory (EPFL LiGHT) with a named team, Swiss AI Initiative and CSCS funding, a contact address, a community tab and commits through 2026-11-17.
What this means for adoptionOwnership is partial. The release is an inspectable medical fine-tune that can be run, adapted and audited on the deployer's own hardware, and the pipeline behind it is public. It is not a clinically usable or clearly licensed product: the publisher bars patient decisions, there is no LICENSE file, the corpus licence is stated two ways, and safety evaluation is thin. Ownership moves to substantial if EPFL ships a LICENSE file and reconciles the corpus licence, and a deployer adds an independent clinical safety evaluation.
Sources
The same evidence records as the entry sheet. Read means the text was verified; unverified means it is known to exist but not yet read.
Model cardread2026-11-19
The Apertus-70B-MeditronFO model card declares licence tag apache-2.0 and an Apache 2.0 link in the header.
Model cardread2026-11-19
The Apertus-8B-MeditronFO model card (partial extracts) has the same template and licence tag, base swiss-ai/Apertus-8B-Instruct-2509, a five-benchmark average of 48.26 against 42.39 for the base, and preference over the base in 74.2 percent of AutoMOOVE comparisons.
Documentationread2026-11-19
Hugging Face records show both repositories are ungated, safetensors only (30 shards for 70B, 4 for 8B, each with an LFS sha256) and without a LICENSE file (raw LICENSE is HTTP 404).
Licenceread2026-11-19
The GitHub code repository EPFLiGHT/FullyOpenMeditron carries an Apache License 2.0 LICENSE file added 2026-08-03.
Documentationread2026-11-19
The project page states the MeditronFO models are not medical devices, have not been tested in clinical practice and must not be used to make decisions about a patient.
Documentationread2026-11-19
The GitHub README says the repository holds the data pipeline, training configs and evaluation to reproduce the paper.
Documentationread2026-11-19
The dataset card gives an Apache-2.0 tag, 533,888 rows in four components and every answer by gpt-oss-120b.
Technical_reportread2026-11-19
arXiv 2605.16215 v2 (29 May 2026) says the corpus is released under a research-use licence and that assets are restricted to research purposes.
Model cardread2026-11-19
The Apertus-70B-Instruct-2509 base card (the 8B card is identical apart from the name) declares licence tag apache-2.0.
Model cardread2026-11-19
gpt-oss-120b, the teacher that wrote every training answer, is Apache-2.0 with a one-sentence usage policy requiring compliance with applicable law.
Vendor announcementread2026-11-19
The EPFL AI Center news item of 17 June 2026 describes MeditronFO as the first fully open framework for medical LLMs and says every stage of development is publicly available.
Vendor announcementread2026-11-19
The Meditron website gives a family history from Meditron 1 (2023) to MeditronFO (2026), a contact address for the lab, and a chat disclaimer saying it is a research model, not a doctor, that can be wrong.
Documentationread2026-11-19
The Hugging Face listing of the EPFLiGHT organisation shows five MeditronFO checkpoints with the licence tag apache-2.0 (Apertus 70B and 8B, OLMo-2-32B, EuroLLM-22B and 9B).
Third-party analysisread2026-11-19
A Hugging Face search for MeditronFO lists community GGUF and i1-GGUF conversions of both Apertus sizes by mradermacher, an MLX conversion of the 8B by loleg and Q4 and Q5 GGUF of the 8B by BrazosDeDios.
Documentationunverified2026-11-19
The Apertus card links an EU AI Act public summary and a Code of Practice document for the base model.