GPT-4.1 API token prices
OpenAI’s published standard API rates for GPT-4.1 (gpt-4.1) are $2.00 per 1 million input tokens, $0.50 per 1 million cached input tokens, and $8.00 per 1 million output tokens, in USD. These are token-usage prices, not a monthly subscription price. OpenAI’s model page identifies the model as gpt-4.1 and lists the same rates; the pricing page lists them under per-million-token pricing. (developers.openai.com)
| Token category | GPT-4.1 rate |
|---|---|
| Input (non-cached) | $2.00 / 1M tokens |
| Cached input | $0.50 / 1M tokens |
| Output | $8.00 / 1M tokens |
Rates shown in OpenAI’s API documentation when consulted on November 12, 2026. Your estimate depends on how many tokens are billed in each category. Count cached tokens at the cached-input rate only when they are reported as cached; do not assume all repeated text qualifies. (developers.openai.com)
Availability caveat: OpenAI’s model page currently marks GPT-4.1 as deprecated. A listed price does not by itself guarantee the model will remain available for a new or ongoing production deployment. Before committing to a production budget, check the model’s current availability and any shutdown_date using OpenAI’s model endpoint; OpenAI says that field can be used to monitor model shutdowns. (developers.openai.com)
Estimate your GPT-4.1 API cost
Use this calculation for token usage. The input rate applies to the non-cached input portion; count tokens billed as cached separately, not again as standard input. If you start with total input tokens, calculate non-cached input = total input − input billed as cached.
Estimated token charge = (non-cached input tokens ÷ 1,000,000 × $2.00) + (cached input tokens ÷ 1,000,000 × $0.50) + (output tokens ÷ 1,000,000 × $8.00).
For example, suppose a workload uses 100,000 standard (non-cached) input tokens and 20,000 output tokens, with no cached input. The estimate is:
- Input: 100,000 ÷ 1,000,000 × $2.00 = $0.20
- Output: 20,000 ÷ 1,000,000 × $8.00 = $0.16
- Estimated total: $0.36
A second hypothetical shows how the cached-input rate changes the math. If a request has 100,000 total input tokens, of which 60,000 are billed as cached, then the remaining 40,000 are non-cached (standard) input tokens. With 20,000 output tokens, the estimate is $0.08 + $0.03 + $0.16 = $0.27. This assumes the 60,000 tokens are actually billed as cached input; it is an estimate, not a guaranteed invoice. The rates come from OpenAI’s published GPT-4.1 pricing. (developers.openai.com)
For a monthly budget, multiply the estimated per-request token charge by the expected number of requests, using realistic average input and output volumes. If usage varies widely, estimate separate low-, typical-, and high-usage scenarios rather than relying on one average.
Which GPT-4.1 price applies to API usage?
These rates are for OpenAI API usage. OpenAI said GPT-4.1 would be available via the API only; it is not a ChatGPT subscription rate. A ChatGPT plan and API usage are different products, so do not use a ChatGPT monthly plan price as the per-token cost of an API workload. (openai.com)
The calculation above covers GPT-4.1 token charges at the listed rates. If your application uses other separately priced services, include those charges separately in the budget. And because OpenAI marks GPT-4.1 deprecated, verify model availability and shutdown information before treating this estimate as a long-term production forecast. (developers.openai.com)