A field guide to medical AI that actually stays current

What it does Awesome-AI4Med is a curated index of resources at the intersection of medicine and artificial intelligence. It catalogs medical large language models, multimodal systems, training datasets, and evaluation benchmarks in structured tables with direct links to papers, model weights, data, and code. The maintainers claim it is automatically extracted from research papers and open-source projects, which explains how it stays current with a field that moves faster than clinical trial recruitment.

The interesting bit The auto-extraction angle is what separates this from a static wiki page. Medical AI has a replication crisis’s worth of models—HuatuoGPT, MedGemma, Baichuan-M3 at 235B parameters—and manually tracking them is a full-time job. The repository also maintains a separate specialized-disease branch, suggesting the curators understand that oncology and psychiatry LLMs should not be lumped together like hospital billing codes.

Key highlights

• Covers both text-only medical LLMs and vision-language variants (HuatuoGPT-Vision, Med-Flamingo, etc.)
• Each entry links to arXiv papers, HuggingFace weights, datasets, and GitHub repos where available
• Includes non-English models, notably Chinese medical LLMs like ShizhenGPT with traditional-medicine training data
• Auto-updated from “a large corpus of research papers and open-source projects” per the README
• Companion repo for disease-specific models exists at Awesome-Specialized-Medical-LLMs

Caveats

• Some entries have missing links: OpenBioLLM and Baichuan-M3 show no paper or data links, and AntAngelMed lacks a paper citation entirely
• The auto-extraction claim is stated but not demonstrated; there is no visible pipeline, commit history detail, or methodology explaining how the automation works
• README was truncated in the provided source, so coverage of datasets and benchmarks sections is incomplete

Verdict Worth bookmarking if you are building or evaluating medical AI systems and need a quick reference to what exists. Skip it if you are looking for original code, reproducible experiments, or critical analysis of whether any of these models actually help patients—the curation is broad, not deep.