OpenAI API Pricing

OpenAI prices its GPT-6 family in three clear tiers. GPT-6 Luna is the high-volume tier at $0.10 per million input tokens, built for classification, extraction and chat workloads where cost per call matters more than peak reasoning. GPT-6.1 Sol is the mainstream tier at $2 input / $10 output, and GPT-6 Astra is the frontier tier at $10 / $50 for the hardest agentic and coding work. Every tier shares the same 1.05M-token context window, so moving between them is a price decision, not an architecture decision.

Two billing details catch teams out. First, OpenAI sells speed as separate service tiers — Fast and Ultrafast multiply the standard rates (Ultrafast Astra is $60 / $300), and those tiers are easy to enable accidentally in a console. Second, prompts above roughly 272K input tokens enter a long-context pricing segment where input and cached rates roughly double; a workload that is cheap in testing can quietly cost twice as much in production once conversations grow. Cached input is the counterweight: Sol reads cached input at $0.10 per million, a 95% discount that rewards stable system prompts.

This page tracks OpenAI’s official API list prices only. OpenRouter and other gateways resell the same models at their own rates, and ChatGPT subscriptions are a separate product with separate limits — neither is mixed into the table below. Every figure is checked against OpenAI’s pricing page and stamped with its verification date.