BioGPT
GPT-style model pretrained from scratch on 15M PubMed abstracts, MIT licensed.
About
BioGPT is a domain-specific generative transformer from Microsoft Research, pretrained from scratch on roughly 15 million PubMed abstracts rather than adapted from a general model. The GPT-2-style decoder comes in a 347-million-parameter base version and a 1.5-billion-parameter BioGPT-Large, and because it learned biomedical language natively it outperformed general models on relation extraction, PubMedQA, and document classification.
Released in October 2022 under the MIT license on GitHub and Hugging Face, it set records of 44.98% F1 on BC5CDR, 38.42% on KD-DTI, 40.76% on DDI, and 78.2% PubMedQA accuracy (81.0% for Large). It has not been updated since the chat-model wave, so treat it as a lightweight, self-hostable specialist for narrow biomedical text tasks rather than a general assistant.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 79; Trending score 70/100 from researcher assessment.