Mage-Flow: An Efficient Native-Resolution Foundation Model for Image Generation and Editing
Microsoft
A compact 4B-parameter native-resolution image generation/editing system pairing Mage-VAE (a single-step-encoding tokenizer cutting tokenization cost by over an order of magnitude versus strong public VAEs) with a diffusion transformer; its Turbo variant does 4-step inference, generating 1024x1024 images in 0.59 seconds on a single GPU while staying competitive on benchmarks.
Why it matters
Shows a small (4B) model can match larger image-generation systems on quality while running an order of magnitude cheaper, relevant to on-device and cost-sensitive image-gen deployment.
Importance: 2/5
Notable but incremental efficiency-focused research release.