GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
fastino
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
fastino
Model Tree
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-BF16
Input Modalities
Output Modalities
Supported Functionality
The model targets biomedical question answering, clinical calculation, medical error detection and correction, biomedical concept, chemical, and disease extraction, and clinical note summarization.
The published weights are BF16 and require about 247 GB before runtime overhead. Tensor parallelism across four 80 GB or larger GPUs (or two B200s) is recommended.
bash
pip install vllm # a release with Nemotron 3.5 Super supportpython
from vllm import LLM, SamplingParams model_id = "fastino/Fastino-Nemotron-3.5-Super-VL-Healthcare" llm = LLM( model=model_id, trust_remote_code=True, dtype="bfloat16", tensor_parallel_size=4, max_model_len=8192,) clinical_note = """65-year-old male with history of hypertension, type 2diabetes, and CKD stage 3 presents with 2 weeks of progressive dyspnea onexertion and orthopnea. Exam: bibasilar crackles, 3+ pitting edema to knees,BP 152/94. Meds: lisinopril 20 mg daily, furosemide 40 mg daily. CXR:cardiomegaly with pulmonary congestion. Assessment: acute decompensatedheart failure.""" messages = [ { "role": "system", "content": "You are a clinical documentation assistant. Summarize notes accurately and concisely. Do not add information that is not in the note.", }, { "role": "user", "content": f"Summarize the following clinical note in 3-5 bullet points:\n\n{clinical_note}", },] outputs = llm.chat( messages, SamplingParams(temperature=0.0, max_tokens=512),)print(outputs[0].outputs[0].text)The data recipe was discovered by the Fastino Fine-Tuning Agent on NVIDIA Nemotron 3.5 Lightning, where it autonomously built evaluation sets, curated data, explored training mixtures and hyperparameters, evaluated transfer, and selected the winning mixture. For Super we ran no new search: the same mixture was trained once on the Super checkpoint through the Fastino Training API.
The adapter was trained on 11,776 de-duplicated examples covering:
It was trained from the base checkpoint for two epochs with LoRA rank 32 on attention, Mamba, latent-projection, and shared-expert weights (75.2M trainable parameters), learning rate 1.6e-4, batch size 32, and sequence packing disabled. Evaluation examples and labels were excluded from training.
Base and fine-tuned scores below use the same early-access checkpoint, inputs, prompts, decoding settings, inference route, and evaluator for each row.
| Benchmark | Evaluation scope | Nemotron base | Fastino-Healthcare | Change |
|---|---|---|---|---|
| PubMedQA | test, n=500 | 70.20% | 77.60% | +7.40 pp |
| MedCalc-Bench Verified | n=1,100 | 65.91% | 72.45% | +6.54 pp |
| MEDEC, flag accuracy | test, n=597 | 72.03% | 79.23% | +7.20 pp |
Performance on related tasks the model was not explicitly trained for.
| Benchmark | Evaluation scope | Nemotron base | Fastino-Healthcare | Change |
|---|---|---|---|---|
| EkaCare | transfer, n=1,066 | 15.85% | 19.70% | +3.85 pp |
| BC5CDR to BioRED | transfer, n=66 documents | 46.07 | 69.04 | +22.97 |
Every benchmark uses a held-out set fixed before training and the same prompts and scorers as our Nemotron 3.5 Lightning study. Score-bearing comparisons were accepted only when base and candidate shared the same examples, prompt construction, decoding parameters, serving route, evaluator identity, and aggregation. Training mixtures were hashed and de-duplicated, and benchmark evaluation rows were excluded from training. Results are from a single training run.
This is a specialized research model, not a medical device. Its outputs are not a substitute for professional clinical judgment. Clinical or other high-stakes deployment requires independent validation, appropriate safeguards, and qualified human oversight.
To fine-tune or serve models like this one without managing GPUs, see the Fastino API, which offers hosted training and inference.
bibtex
@misc{atreja2026pioneeragentcontinualimprovement, title={Pioneer Agent: Continual Improvement of Small Language Models in Production}, author={Dhruv Atreja and Julia White and Nikhil Nayak and Kelton Zhang and Henrijs Princis and George Hurn-Maloney and Ash Lewis and Urchade Zaratiana}, year={2026}, eprint={2604.09791}, archivePrefix={arXiv}, primaryClass={cs.AI}, url={https://arxiv.org/abs/2604.09791},}This model is licensed under the Apache License 2.0.
| MEDEC, sentence accuracy | test, n=597 | 70.02% | 78.56% | +8.54 pp |
| MedMentions ST21pv, span F1 | test, n=879 documents | 40.44 | 50.38 | +9.94 |
| BC5CDR, span + type F1 | test, n=500 documents | 63.82 | 84.80 | +20.98 |