ANX Model Lab
Exploring native AI models, from the first weights to evaluation.
The project
Start at the model itself.
ANX Model Lab investigates model development from randomly initialized weights, with work around training, tokenization, datasets, evaluation, and checkpoints.
It is a research project for exploring architecture and training choices. Production usefulness and independently reproduced performance are not established.
What it explores
- Training and architecture experiments
- Tokenization and dataset workflows
- Evaluation and checkpoints
Where it stands
Research stage. No production-ready model or public model distribution is claimed.
In the ecosystem
Native models could eventually support parts of the ecosystem. The documented HQ integration is explicitly design-only.
Keep exploring