Model reference · open weights

IF-I-XL

IF-I-XL is an open-weight image model from DeepFloyd. IF-I-XL-v1.0 (FP32) weighs 14.3 GB; the smallest configuration that runs it is RTX 4060 Ti 16 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

• IF-I-XL is a 4.3B parameter pixel-based text-to-image diffusion model developed by DeepFloyd and StabilityAI.
• It functions as the first stage of a triple-cascaded system that generates 64x64 images using a frozen T5 text encoder and a UNet architecture.
• The model is primarily designed for English prompts and is released under the DeepFloyd IF License Agreement.

Summary of the DeepFloyd/IF-I-XL-v1.0 model card, 2026-10-01

Hugging Face ↗Paper ↗