FLUX 1 [Schnell] (12 B params, Apache-2.0)

Fast-track text-to-image model distilled for speed.

• 1-to-4-step renders. Rectified-flow backbone + latent adversarial distillation means high-quality 1024-px frames in as few as four denoise loops (one if you crank the speed).

• Looks like the big dogs. Prompt following and image fidelity rival closed-source heavyweights while staying fully open-weight.

• Fits modest cards. Runs comfortably on an 8 GB GPU; with CPU offload it even squeezes onto a 4 GB mobile 1050—just slower.

• Plug-and-play tooling. Native FluxPipeline in Diffusers, ComfyUI node, and ready-made ONNX (BF16/FP8/FP4) checkpoints—from_pretrained("black-forest-labs/FLUX.1-schnell") and go.

Why pick it for Norman AI?

SDXL-level visuals at SSD-level latency, open licensing, and single-card footprints let us roll out instant hero art, real-time previews, or an edge “speed” tier without burning H100 hours.

response = await norman.invoke(
    {
        "model_name": "stable-diffusion-v1-5",
        "inputs": [
            {
                "display_title": "Prompt",
                "data": "A cat playing with a ball on mars"
            }
        ]
    }
)