Train LoRAs for Qwen-Image 2.1, FLUX.2 Klein 9B, Krea 2, Z-Image, Ideogram 4, Anima, SDXL (Pony, Illustrious, NoobAI), LTX 2.3 and MiniMax-H3 on NVIDIA GPUs from 4–8 GB VRAM. One install, one launcher, NF4 models. Windows, Linux and RunPod.
End-to-end benchmark for AI-generated GPU kernels, drawn from real production traces — turn a PyTorch reference into a DSL kernel (Triton, Gluon, FlyDSL, CuteDSL) and grade it on compilation, numerical correctness, and speed-of-light efficiency, on both AMD and NVIDIA GPUs.