# Atlas Inference

Pure Rust + CUDA LLM inference engine, hand-tuned for prosumer AI machines (NVIDIA DGX Spark / GB10 and AMD Strix Halo / gfx1151). Submitted to MLPerf Inference v6.1, closed edge division, on both.

## Links

- Repo: https://github.com/Avarok-Cybersecurity/atlas
- MLPerf Edge Agentic benchmark (Atlas Inference named as a contributor): https://mlcommons.org/2026/07/mlperf-inference-v61-edge-agentic/
- Deployment guide: https://github.com/Avarok-Cybersecurity/atlas/blob/main/docs/GB10_DEPLOYMENT_GUIDE.md
- Recipes (model SSOT): https://github.com/Avarok-Cybersecurity/atlas-recipes
- Discord: https://discord.gg/6vDbKaKrKD
- Site: https://atlasinference.io