Alpha
Contact
Use · MeditronFO

Is it good enough?

Ownership levelPartialnone·limited·partial·substantial·fullAnalytical input D ยท 51.2/100

This page is a projection of the one entry record, the Reliability factor that Use 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

Capabilities & modalities

Text in, text out, English medical content. The fine-tuning corpus covers four kinds of material:

  • public medical exam and patient-question datasets (MedMCQA, MedQA, AfriMed-QA, HealthSearchQA, LiveQA-Med);
  • synthetic exam-style questions;
  • questions generated from 14,192 clinical practice guidelines;
  • open-ended adversarial clinical vignettes modelled on clinician-written MOOVE cases (ev-data-card).

Compared with its Apertus-Instruct base, the 70B gains 9.2 points on MedQA, 7.4 on MedMCQA, 4.3 on MedXpertQA, 9.9 on HealthBench and 0.2 on PubMedQA (ev-card-70b). The 8B gains on MedQA and MedMCQA but loses 2.9 points on PubMedQA (ev-card-8b). It is weaker at general instruction following than its base (ev-project-page). There is no image or audio input.

Context window & long-context behaviour

config.json sets 65,536 positions with llama3-style rope scaling (factor 8 from an original 8,192) for both sizes (ev-hf-meta). The fine-tuning packed sequences of 8,192 tokens, and the card's generation default is 8,192 new tokens (ev-card-70b). The publisher reports no long-context evaluation of the fine-tune, so the base model's long-context quality may not have survived. Test clinical records, long guidelines and multi-document prompts against the deployer's own cases first.

Prompt format & chat template

Both repositories ship a chat_template.jinja (ev-hf-meta). Use tokenizer.apply_chat_template(messages, add_generation_prompt=True, ...) as the card does, with a single user turn and no system prompt. The AutoMOOVE answering setup also used the vignette alone with no system prompt (ev-card-70b, ev-github-readme). Sample with temperature 0.7 as in the card and generation config, which is what the benchmark numbers used for open-ended tasks. The multiple-choice benchmarks were decoded greedily (ev-card-70b).

Language coverage

The cards tag English only, and the corpus is English (ev-card-70b, ev-data-card). The Apertus base is advertised as multilingual with support for over 1,000 languages (ev-base-card). The paper and cards report no multilingual medical evaluation, so any non-English clinical use is untested.

How this scores

The ownership factor this domain covers, drawn from the one entry record.

3

ReliabilityIs it reliable and good enough for the job?

Weak

Safety is 2: no safety tuning, publisher-only low-agreement safety judging and no guard or red-team, and the publisher says not to use it for patient decisions, so it is not safe to deploy as presented.

How this scores (AOI sub-dimensions)
Operational3/5how practical it is to run, serve and maintain in productionStandard transformers architecture class supported since transformers 4.56 per the base card, BF16 safetensors, a chat template and recommended sampling settings in generation_config, and community GGUF and MLX conversions for both sizes.
Safety2/5whether misuse risks are evaluated and guardrails are providedThe models are supervised fine-tunes for medical accuracy and not safety tuning.
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.