Gelato-30B-A3B-f32-AIO-GGUF

Gelato-30B-A3B-f32-AIO-GGUF

Gelato-30B-A3B is a 30B-parameter Qwen3-VL MoE–based grounding model specialized for GUI computer-use tasks, trained on the Click-100k dataset to map natural language instructions and screen images to precise click coordinates on user interfaces. It achieves state-of-the-art accuracy on key grounding benchmarks, reaching about 63.88%63.88% on ScreenSpot-Pro and 69.15%/74.65%69.15%/74.65% on OS-World-G / OS-World-G (Refined), outperforming prior dedicated computer grounding models such as GTA1-32B and even larger general-purpose VLMs like Qwen3-VL-235B-A22B-Instruct. The model is released with an open codebase and examples showing how to feed a GUI screenshot plus an instruction and obtain normalized (x,y)(x,y) coordinates, making it a strong drop-in component for building computer-use agents that can reliably locate UI elements and interact with real software environments.

Model Files

Quants Usage

(sorted by size, not necessarily quality. IQ-quants are often preferable over similar sized non-IQ quants)

Here is a handy graph by ikawrakow comparing some lower-quality quant types (lower is better):