Huawei Hong Kong AI Lab Open Sources 10B Parameter Boogu Image Generation Model

Huawei releases Boogu-Image-0.1, a 10B parameter model trained for $400k on 2B images. It supports 2K resolution and multilingual generation under Apache 2.0.

Woofun AI reports that Huawei's Hong Kong AI Lab, in collaboration with university researchers, has open-sourced Boogu-Image-0.1. This 10 billion parameter image generation and editing model includes Base, Turbo, Edit, and Edit-Turbo variants, supporting both Chinese and English text prompts.

The Base model was trained from scratch using 2.0862 billion de-duplicated images, with a theoretical training cost of approximately $400,000. The paper states it reaches the top tier of open-source models on standard benchmarks, with some metrics approaching leading closed-source alternatives. Base and Edit models support up to 2K resolution, while Turbo generates 1K images in 3 to 4 steps. The system rephrases user prompts and selects models based on task difficulty to minimize inference costs. Weights, inference code, and training methods are released under the Apache 2.0 license.