MiniMax has released H3-Image, an open-weight model with 33.1 billion parameters available on Hugging Face, through MiniMax Code and via the company’s API. The model combines text-to-image generation and image editing in a diffusion transformer architecture, with native output of approximately 16 megapixels, reaching 4096×4096 pixels or 6272×2688 pixels. It requires about six sampling steps, accepts up to 16 reference images and relies on a separate Qwen3-VL-32B text encoder for local use. The API costs approximately $0.022 per 1K image and $0.069 per 4K image, while the community license prohibits use in the US, EU, UK and South Korea. Large companies also need written authorization from MiniMax, and the derivative model P-H3-Image was announced on November 3, 2026, with claims of faster inference on consumer hardware.
Source: Read the original article
