Pruna AI's P-H3-Image brings text-to-image generation and multi-image editing to a model derived from MiniMax H3. Its November 3 release includes a Diffusers pipeline and a separate ComfyUI package. This report was published November 8.
The license comes first
The model card says the MiniMax H3 Community License excludes the United States, European Union, United Kingdom and Republic of Korea unless a separate license is obtained. It also identifies a product attribution condition. Downloadable weights do not make this an unrestricted open-source model; read the actual license and notices for your intended use.
Hardware and quality
Pruna says its quantized ComfyUI files run in 21 GB of VRAM in the configuration described. That is above a 12 GB card, and offloading behavior and performance need separate testing. The card describes eight-step generation and warns that small text and details can fail.
For our readers, the geographic restriction is more important than the headline speed: a model is only a viable workflow if its terms permit the use and location. We have not downloaded, run or evaluated this model.
Original reporting; claims attributed to their sources. No local tests implied.
