Linarix-v1 — Text to Image

Generate 1024×1024 images from text with a ~712M GatedDeltaNet-2 (linear-attention) diffusion model. Strong on landscapes, scenic and man-made structures, and composed portraits — use descriptive prompts; CFG 4.0 is the sweet spot. For exact object counts, try a few seeds. Runs on ZeroGPU (NVIDIA RTX Pro 6000 Blackwell, 48 GB). First run may take a minute while a GPU is allocated.