RUN IT YOURSELF OR USE A HOST

Rent a GPU or pay per token?

Output tokens per month, in millions

300

API price per million output tokens

GPUs you would rent, running all month

Rental price per GPU-hour

Monthly API bill$3,000/mo

Monthly GPU rental$2,912.7/mo

Rental = GPUs x hourly price x 730 hours. Ignores input tokens, setup, idle time and engineer time. Defaults: Sonnet 5.5 output price, Lambda 80 GB GPU rate.