Best way to test models at different quants before buying GPUs

A Reddit user is seeking advice on the most effective method for testing model performance across various quantization levels prior to purchasing new hardware.

• The user intends to buy multiple GPUs and specifically wants to benchmark models like GLM-5.2.
• The proposed solution involves renting RTX 6000 GPUs via the Vast.ai platform to run these tests.