Pre-trained base models without chat-style instruction tuning, ideal for fine-tuning and continued pre-training.
Foundation models (a.k.a. base models) are pre-trained on large unlabeled corpora and retain the full language-modeling capability, but they are not directly optimized for chat. They are best suited for downstream fine-tuning, continued pre-training, or as research baselines. The list below covers parameter sizes, context windows, licenses, and download links for mainstream base models.
Filter by type, size, license, or publisher to narrow down the list.
0 models cataloged
Loading models...
