MOSS-Image Passes 1.3 Million Downloads in Four Days, Tops Hugging Face Trends

MOSS-Image, the 5B-parameter open-source image generation model from MOSI and OpenMOSS, drew about 1.3 million downloads in its first four days and reached the top of Hugging Face's text-to-image trending list. It pairs a 5B diffusion transformer with a frozen Qwen3.5-9B text encoder and the FLUX.2 VAE, and supports dense Chinese and English text rendering plus prompts up to 4,096 tokens.

• About 1.3 million downloads in four days across several repositories
• 5B diffusion transformer with Qwen3.5-9B encoder and FLUX.2 VAE
• Prompts up to 4,096 tokens and JSON layout input with bounding boxes
• Community shipped 4-step and 8-step versions, INT8, GGUF and MLX builds