GPT-4o API pricing at a glance
As checked on November 9, 2026, OpenAI lists the gpt-4o API alias at $2.50 per 1 million input text tokens and $10.00 per 1 million output text tokens under its Standard pricing. Eligible cached input is listed at $1.25 per 1 million tokens. The pricing page presents Batch pricing separately; the rates and example below are for Standard processing, not Batch. OpenAI’s pages do not state an effective date for these specific rates.
These are API token prices in USD—not a ChatGPT subscription price. The calculation below covers text tokens only; it does not estimate image usage or separate tool or storage charges.
How to estimate a text workload
Calculate input and output separately, then add them:
Cost = (input tokens ÷ 1,000,000 × $2.50) + (output tokens ÷ 1,000,000 × $10.00)
For example, suppose a workload uses 100,000 input tokens and 10,000 output tokens per day, all billed at Standard rates with no cached input:
- Input: 100,000 ÷ 1,000,000 × $2.50 = $0.25 per day
- Output: 10,000 ÷ 1,000,000 × $10.00 = $0.10 per day
- Total: $0.35 per day, or $10.50 for 30 days
This is an estimate of token charges for the stated text workload. It excludes any applicable tool or storage charges, image usage, taxes, third-party access fees, and other charges not represented in the calculation. For eligible cached input, apply the listed cached-input rate to those tokens instead: 100,000 cached input tokens at $1.25 per million would cost $0.125.
What this price covers—and what it doesn’t
The quoted rates apply to OpenAI API token usage for the gpt-4o alias, not ChatGPT subscription access. OpenAI bills API usage separately from ChatGPT subscriptions. If you use a third-party provider to access the API, account for any fees or conditions that provider adds separately.
OpenAI’s model page describes GPT-4o as accepting text and image inputs and producing text outputs. This estimate covers text-token usage only. The model page labels GPT-4o deprecated, while OpenAI’s Help Center says it remains available through the API and that OpenAI will provide advance notice before future API retirements. The page also identifies the older snapshot gpt-4o-2024-05-13 as shut down. For a production deployment, distinguish the gpt-4o alias from that retired snapshot and check the current model documentation before relying on a specific model ID.