Docto Decision Qwen3.5-4B (v0.1)
French medical decision models by Bofeng Huang: pass a patient message or clinical note, a question and candidate answers, and get one probability per answer from a single forward pass. Choice, score and noul. A research model, not a medical device.
A rank-16 LoRA merged into Qwen3.5-4B; the answer is one token (an option letter, or oui/non for noul). French only, up to 26 options. A research model, not validated clinically, and it can under-estimate urgency. Most benchmark labels come from Qwen teachers; on the gold-labelled tasks it scores 70.13% (ICD-10 chapter) and 60.94% (article specialty). Siblings: ModernBERT-fr-base (150M), Qwen3.5-0.8B and 2B, Gemma-4 E2B and E4B (all Apache-2.0) and MedGemma-1.5-4B (Google HAI-DEF terms). Trained on docto-decision-data-fr-v0.1 (943,013 records, 108 tasks).
What it decides
- choice — picks one option from a set
- score — places the input on an ordered scale
- noul — answers a yes/no question with one probability
At a glance
| Parameters | 4.66B |
| Base model | Qwen/Qwen3.5-4B |
| Maker | Bofeng Huang |
| Released | 2026-10-02 |
| License | apache-2.0 |
| Reported accuracy | 88.9% |
| Reported latency | 47 ms median per request on one H100 in BF16 |
Get the weights
pip install systemonemodels
systemone pull bofeng-huang/docto-decision
The files are served from the maker's Hugging Face repository, bofenghuang/docto-decision-qwen3.5-4b-fr-v0.1, and verified against the checksums recorded here.
Read more
This page was opened by System One for Bofeng Huang, who can claim the organisation and take it over at any time.