Marktechpost AI has introduced the GPU Leaderboard, a pricing dashboard designed to compare costs of rentable cloud GPUs. The tool allows teams to calculate expenses based on specific accelerators, GPU counts, duration, and utilization rates. It provides a comparative analysis across 15 different cloud providers.
For a standard configuration of 8xH100 GPUs over 72 hours at 90% utilization, the platform reports a range of costs. The lowest price identified is $967.68 at Latitude, while the highest reaches $3,545.28 at CoreWeave. The median cost for this specific workload is $1,716.48.
The dashboard also calculates the effective cost per used GPU-hour, accounting for idle time billing. For instance, a base rate of $1.68 per hour at 90% utilization adjusts to $1.87 per hour. This metric is intended to provide more accurate budget forecasting for inference tasks.
Alongside the leaderboard, the company released the Deploy It Chrome extension. This tool integrates pricing data directly into developer workflows, allowing users to access cost information while selecting models. The extension is currently available for installation via the Chrome Web Store.
