Skip to main content
Usage helps you understand the research loop’s resource footprint. It can show evaluation runs, CPU time, memory time, aggregate GPU device time, Droyd sandbox charges, and model-token activity by the selected period or competition. GPU cost is shown as one total even when runs use different chip types. LLM token information is useful for debugging and model selection. Your OpenRouter account pays for that inference directly, so Droyd shows no separate LLM charge.

​
Standard sandbox examples

These examples use the standard 2-core, 4-GiB sandbox. See the exact resource formula on the pricing page.