Nucleotide Transformer

DNA foundation language model family (500M-2.5B params) pretrained on 850+ genomes, including the human 1000 Genomes collection.

Unsigned

Overview

The Nucleotide Transformers are a collection of foundational language models pretrained on DNA sequences from whole genomes across 850+ species, developed by InstaDeep in collaboration with NVIDIA and the Technical University of Munich. The models integrate information from multiple species during pretraining and are widely used as a base for downstream genomic tasks such as variant-effect and regulatory-element prediction.

**Registry note on `model_card_uri`**: this points to the model's standard Hugging Face model card, not a GA4GH Genomic AI Model Card v0.6.0-formatted document. No model has yet published a card in the GA4GH v0.6.0 format, since it is a newly drafted specification; this entry is transparent about that gap rather than implying conformance that does not exist.

Details

Licence: other (full text)
Version: 2.5b-multi-species
Category: Genomic foundation model
Homepage: https://huggingface.co/InstaDeepAI/nucleotide-transformer-2.5b-multi-species
Repository: https://github.com/instadeepai/nucleotide-transformer
Maintainers: Susheel Varma (@susheel, GA4GH AI Workstream / Sage Bionetworks)
Keywords: dnagenomicsfoundation-modelmulti-specieslanguage-model

Safety classification

No safety classification has been submitted for this entry.

Model Card

View the full GA4GH Genomic AI Model Card ↗

Hosting: huggingface
Access: https://huggingface.co/InstaDeepAI/nucleotide-transformer-2.5b-multi-species

Mirrored data last synced from the Model Card at 2026-09-04T11:05:00.000Z. See the linked card for the authoritative, live version.

Provenance

Created: 2026-09-04T11:05:00.000Z
Updated: 2026-09-04T11:05:00.000Z
Last verified: 2026-09-04T11:05:00.000Z