NeuroBrix distinguishes four levels of hardware support. A vendor being detected is not the same as being supported.
| Level | Meaning | Hardware |
|---|---|---|
| Validated | Models run on it in the stable engine | NVIDIA GPUs (both engines), Apple Silicon (the Triton engine through Metal, the PyTorch engine through MPS), CPU (the PyTorch engine) |
| Experimental | Tried by the engine, not yet validated on hardware by the NeuroBrix lab | Intel GPUs with their Level Zero runtime |
| A target | Code exists, not yet validated on a card | AMD GPUs, through nbxtensor's HIP driver |
| Recognised only | Detection identifies it; nothing executes | Tenstorrent, Moore Threads, Biren, Iluvatar, Hygon DCU, Cambricon |
See Hardware Support for details.
Run neurobrix autotune certify on your card and open a pull request with the certificate template. Every GPU class you certify is served to everyone who has that card.
Submit a YAML hardware profile for your machine. Format: Hardware Profiles.
Run published models on your hardware and report results and failures in GitHub Issues. Include the output of neurobrix doctor.
Whether a formal verification programme or badge exists: not documented yet. Hardware partnerships: [email protected].
Next: Hardware Support
© 2025–2026 Hocine Benkelaya. NeuroBrix engine released under the Apache License 2.0.
Developed by Hocine Benkelaya · Supported and funded by Neural Network Holding Ltd · Commercial distribution by WizWorks OÜ
Documentation source: GitLab