Fibo-2.0-dev: New 8.1B-Parameter text-to-image Model Released on HuggingFace

On October 18, 2026, the model briaai/Fibo-2.0-dev for text-to-image generation was published on HuggingFace. In its first days the model gathered 35 downloads and 69 likes.

What it means

The model contains 8.1 billion parameters, the largest weights file takes up 9.3 GB, and the entire repository is 23.7 GB in safetensors format. It runs via the diffusers library. License — CC BY-NC 4.0: the model may be used for non-commercial purposes only, with mandatory attribution. Copying, distributing and remixing the material is allowed in any format, but you may not add restrictions that prevent others from doing what the license permits. Running it locally will require a GPU with enough memory for 23.7 GB of weights, or the use of quantized builds if they appear. The developer does not specify minimum VRAM requirements and does not publish ready-made quantized versions in the repository. Those looking for a text-to-image model for running on your own hardware should keep in mind the non-commercial license and the considerable size of the weights. You can follow news about AI tools for 3D and video in the CPU3D news feed.