Fastino Nemotron 3.5 Super VL Healthcare

Users can run it for biomedical question answering, clinical calculations, medical error detection and correction, biomedical information extraction, and clinical note summarization.

Availability

Published weights are available on Hugging Face.

Pricing model

Not established

Deployment

BF16 weights; about 247 GB required before runtime overhead.

Privacy

Not established

Licence

Apache License 2.0

• The model targets biomedical question answering, clinical calculation, medical error detection and correction, biomedical extraction, and clinical note summarization. ^[1]

  The model targets biomedical question answering, clinical calculation, medical error detection and correction, biomedical concept, chemical, and disease extraction, and clinical note summarization.

• The repository contains a ready-to-run merged checkpoint, so users do not need to obtain or attach a separate LoRA. ^[1]

  Users do not need to obtain or attach a separate LoRA.

• The published BF16 weights require about 247 GB before runtime overhead. ^[1]

  The published weights are BF16 and require about 247 GB before runtime overhead.

• The model is a specialized research model, not a medical device, and its outputs are not a substitute for professional clinical judgment. ^[1]

  Its outputs are not a substitute for professional clinical judgment.

• On PubMedQA, the reported score was 77.60%, compared with 70.20% for the base checkpoint. ^[1]

  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.

• DeepSeek opens a real-time web search API Weekly brief 2026-W42
• DeepSeek adds web search for apps and updates its Flash model Daily brief 2026-10-17