According to CoinMeta, Bria has released a new image model, Fibo 2.0, which boasts 8 billion parameters for the first time integrating text-to-image generation and image editing into a single model. Users can modify photos through natural language, or use structured JSON to specify locations of people, lighting, and text, avoiding damage to other parts of the image. Fibo 2.0 supports mask editing and multi-image reference, and also offers a full development version as well as an accelerated image editing Edit Turbo. Based on the artificial analysis's AA-image-editing v2.1 rankings, Fibo 2.0 ranks first among open-source models with 1087 ELO, slightly ahead of Qwen-image-2.1's 1083. In the text-to-image generation rankings, Qwen leads with 1042 points, surpassing Fibo which has 1039 points, with a difference that is not statistically significant. Fibo 2.0 is trained using licensed images, and Bria collaborates with copyright holders such as Getty Images to obtain materials. The model's weights are licensed under the CC BY - NC 4.0 non-commercial license; commercial use requires separate contracting.