Domain models
Models for medicine, law, finance and science: Med-PaLM, BloombergGPT, Galactica. 23 models on EveryLLM so far.
Domain models are language models built or tuned for one field, such as medicine, law, finance, science or robotics. Some are trained from scratch on a field's own text; more often a general model is fine-tuned on specialist data and tested against the field's own examinations. Their promise is better accuracy and vocabulary in a narrow area. Their risk is that confident errors matter more when a diagnosis or a contract depends on them.
Medicine
Google's Med-PaLM, described in December 2022, was the first model reported to pass the mark used for US medical licensing exam questions, scoring 67.6% on the MedQA dataset; Med-PaLM 2 reached 86.5% in 2023. Google later released MedGemma (2025) as open weights. OpenEvidence, a clinical answer service for doctors, is built on language models that cite the medical literature. Most clinical tools are offered as decision support, and regulators treat them with caution.
Finance, science and law
BloombergGPT (March 2023) was a 50-billion-parameter model trained on a mix of financial and general text. Meta's Galactica, a 120-billion-parameter science model, was withdrawn from public demo three days after its launch in November 2022 after it produced plausible but wrong scientific text. Legal tools are mostly general models tuned and grounded on case law, and several courts have sanctioned lawyers who filed briefs with invented citations.
Famous domain models
Still to be written: OpenEvidence.