Arioron Labs
Cutting-edge AI research lab creating the next generation of intelligent systems — including the VexAI assistant and advanced language models published openly for anyone to check.
Visit- Founded
- 2024
- Published models
- 6
- Open datasets
- 3
- Largest model
- 2B parameters
- Licence
- Open weights
Constraint as a research method. We are more interested in what two billion parameters can do than in what two hundred billion can.
Intelligence amplified.
Arioron Labs does applied research with a bias toward constraint. The frontier is well funded and well staffed; we are more interested in what can be achieved with two billion parameters than with two hundred.
Two lines of work dominate. The first is parameter efficiency in small language models — how much reasoning and coding capability can be recovered at a size that runs on hardware an ordinary organisation already owns. The second is medical imaging, where we work on diagnostic models built to survive real hospital data rather than curated benchmark sets.
We publish weights, datasets and evaluation code so the claims can be checked.
- Small language model research
- Parameter efficiency and distillation
- Medical imaging and diagnostic models
- Evaluation methodology
- Open weights and datasets
Research
Open roles
Positions within Arioron Labs.