If you’re looking for fully FOSS LLM models - not just open-weight ones, but those that also include training scripts, datasets, and checkpoints - here’s a list of some of the recent:

- SmolLM2, a series of strong small models
- SmolLM3, SOTA 3B model with dual reasoning, supports 6 languages and long context with strong function calling
- OLMo-7B/1B, slightly older models with shorter context size
- Pythia suite, a collection of 8 models with parameters ranging from 70M to 12B