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

File Name	Quant Type	File Size
Gelato-30B-A3B-BF16.gguf	BF16	61.1 GB
Gelato-30B-A3B-F16.gguf	F16	61.1 GB
Gelato-30B-A3B-F32.gguf	F32	122 GB
Gelato-30B-A3B.IQ4_XS.gguf	IQ4_XS	16.6 GB
Gelato-30B-A3B.Q2_K.gguf	Q2_K	11.3 GB
Gelato-30B-A3B.Q3_K_L.gguf	Q3_K_L	15.9 GB
Gelato-30B-A3B.Q3_K_M.gguf	Q3_K_M	14.7 GB
Gelato-30B-A3B.Q3_K_S.gguf	Q3_K_S	13.3 GB
Gelato-30B-A3B.Q4_K_M.gguf	Q4_K_M	18.6 GB
Gelato-30B-A3B.Q4_K_S.gguf	Q4_K_S	17.5 GB
Gelato-30B-A3B.Q5_K_M.gguf	Q5_K_M	21.7 GB
Gelato-30B-A3B.Q5_K_S.gguf	Q5_K_S	21.1 GB
Gelato-30B-A3B.Q6_K.gguf	Q6_K	25.1 GB
Gelato-30B-A3B.Q8_0.gguf	Q8_0	32.5 GB
Gelato-30B-A3B.i1-IQ1_M.gguf	i1-IQ1_M	7.08 GB
Gelato-30B-A3B.i1-IQ1_S.gguf	i1-IQ1_S	6.42 GB
Gelato-30B-A3B.i1-IQ2_M.gguf	i1-IQ2_M	10.2 GB
Gelato-30B-A3B.i1-IQ2_S.gguf	i1-IQ2_S	9.29 GB
Gelato-30B-A3B.i1-IQ2_XS.gguf	i1-IQ2_XS	9.08 GB
Gelato-30B-A3B.i1-IQ2_XXS.gguf	i1-IQ2_XXS	8.18 GB
Gelato-30B-A3B.i1-IQ3_M.gguf	i1-IQ3_M	13.5 GB
Gelato-30B-A3B.i1-IQ3_S.gguf	i1-IQ3_S	13.3 GB
Gelato-30B-A3B.i1-IQ3_XS.gguf	i1-IQ3_XS	12.6 GB
Gelato-30B-A3B.i1-IQ3_XXS.gguf	i1-IQ3_XXS	11.8 GB
Gelato-30B-A3B.i1-IQ4_XS.gguf	i1-IQ4_XS	16.4 GB
Gelato-30B-A3B.i1-Q2_K.gguf	i1-Q2_K	11.3 GB
Gelato-30B-A3B.i1-Q2_K_S.gguf	i1-Q2_K_S	10.5 GB
Gelato-30B-A3B.i1-Q3_K_L.gguf	i1-Q3_K_L	15.9 GB
Gelato-30B-A3B.i1-Q3_K_M.gguf	i1-Q3_K_M	14.7 GB
Gelato-30B-A3B.i1-Q3_K_S.gguf	i1-Q3_K_S	13.3 GB
Gelato-30B-A3B.i1-Q4_0.gguf	i1-Q4_0	17.4 GB
Gelato-30B-A3B.i1-Q4_1.gguf	i1-Q4_1	19.2 GB
Gelato-30B-A3B.i1-Q4_K_M.gguf	i1-Q4_K_M	18.6 GB
Gelato-30B-A3B.i1-Q4_K_S.gguf	i1-Q4_K_S	17.5 GB
Gelato-30B-A3B.i1-Q5_K_M.gguf	i1-Q5_K_M	21.7 GB
Gelato-30B-A3B.i1-Q5_K_S.gguf	i1-Q5_K_S	21.1 GB
Gelato-30B-A3B.i1-Q6_K.gguf	i1-Q6_K	25.1 GB
Gelato-30B-A3B-mmproj-bf16.gguf	mmproj-bf16	1.09 GB
Gelato-30B-A3B-mmproj-f16.gguf	mmproj-f16	1.08 GB
Gelato-30B-A3B-mmproj-f32.gguf	mmproj-f32	2.15 GB
Gelato-30B-A3B-mmproj-q8_0.gguf	mmproj-q8_0	712 MB
Gelato-30B-A3B.imatrix.gguf	imatrix	122 MB

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):