Monitor your spend
Pick the right GPU
Delete instances when done
Use volumes to avoid re-setup costs
Re-downloading models wastes 10-30 minutes of GPU time per session:/workspace/ persist across deploys.
Right-size your instance
Batch your work
Deploy, process, tear down: pay only for the minutes your job runs:Use the Models API for light workloads
For light or bursty inference, the Models API bills per token, so you pay nothing while a GPU would sit idle:Quick checklist
- Run
runcrate psdaily: kill anything not in use. - Run
runcrate billing usageweekly: spot unexpected charges early. - Use volumes for models and data: avoid re-downloads.
- Match GPU to workload: check
nvidia-smiutilization. - Delete instances immediately after batch jobs complete.