Billing and pricing

Pre-flight launches are free (see Errors and limits). Paid, always-on launches are opening soon; this page describes how they will be priced, which is also what Pricing shows.

From GPU hour to token price

• Output tokens = GPU $/hr ÷ (200 tokens/s × 3600) × 1,000,000 × 1.5.
• Input tokens cost 25% of output tokens.
• Start fee: what the GPU costs during a cold start (about 130 seconds of its hourly price), charged once when the launch succeeds.

Pilot picks the smallest GPU that fits your model, so a smaller model is cheaper per token. GPUs scale to zero when idle, and a launched model nobody calls for 7 days is deleted. Each model’s own per-million-token prices are returned by GET /v1/models. The model pages show the typical GPU tier and price for each popular model.

More from litepod: the free LLM API with free open-source models. ModelDock is a litepod product.