Direct answer: GPT-5.5 API rates
OpenAI’s API pricing documentation lists standard GPT-5.5 at $5 per 1 million input tokens and $30 per 1 million output tokens for requests below the long-context threshold. Cached input is listed at $0.50 per 1 million tokens. These are USD token rates—not a monthly subscription fee. The GPT-5.5 Pro variant costs more. Rates below reflect the OpenAI documentation available on November 9, 2026. (developers.openai.com)
| Model and context | Input / 1M tokens | Cached input / 1M tokens | Output / 1M tokens |
|---|---|---|---|
| GPT-5.5, under 272K context | $5.00 | $0.50 | $30.00 |
| GPT-5.5, long context | $10.00 | $1.00 | $45.00 |
| GPT-5.5 Pro, under 272K context | $30.00 | Not listed | $180.00 |
| GPT-5.5 Pro, long context | $60.00 | Not listed | $270.00 |
OpenAI says a prompt with more than 272K input tokens is charged at twice the input rate and 1.5 times the output rate for the full session. The rate card presents the corresponding short- and long-context prices above. If your request is near the threshold, check the current rate card and how your request is counted before using a budget estimate. (developers.openai.com)
The GPT-5.5 model page also lists a 10% uplift for regional-processing (data-residency) endpoints. That applies when you use an eligible regional endpoint; it is not a general surcharge on every GPT-5.5 request. (developers.openai.com)
How to calculate GPT-5.5 API cost
For standard GPT-5.5 usage below the long-context threshold, estimate token charges as:
uncached input tokens = total input tokens − cached input tokens
(uncached input tokens ÷ 1,000,000 × input rate)
+ (cached input tokens ÷ 1,000,000 × cached-input rate)
+ (output tokens ÷ 1,000,000 × output rate)
Uncached input and cached input are non-overlapping parts of total input: add them together to get total input, and do not count cached tokens again at the standard input rate. Use zero for cached input if none of your tokens receive the cached rate. Multiply the per-request estimate by the number of similar requests in your billing period. This estimates token charges; applicable tool or storage charges may add to the bill. (developers.openai.com)
Worked example: Suppose one request uses 2,000 input tokens and 500 output tokens, with no cached input. At $5 per million input tokens and $30 per million output tokens:
- Input: 2,000 ÷ 1,000,000 × $5 = $0.01
- Output: 500 ÷ 1,000,000 × $30 = $0.015
- Total: $0.025 per request before rounding
At 1,000 identical requests each day for 30 days, that works out to $750 in estimated token charges. This example assumes the short-context standard rates apply, excludes tool or storage charges, and does not include any regional-processing uplift.
For a cache illustration, consider 1 million aggregate input tokens across multiple requests, with each request individually below the long-context threshold. If 800,000 tokens are cached and 200,000 are uncached, the listed short-context rates give $1.40 in input charges: (200,000 ÷ 1,000,000 × $5) + (800,000 ÷ 1,000,000 × $0.50). This assumes those 800,000 tokens actually qualify for cached-input pricing.
Pricing options and conditions that can change your estimate
- GPT-5.5 Pro: The standard-context rates are $30 per million input tokens and $180 per million output tokens. The long-context rates are $60 and $270, respectively. The rate card does not list a cached-input price for Pro, so do not assume the standard GPT-5.5 cache rate applies. (
developers.openai.com) - Long-context requests: Above 272K input tokens, the higher rates apply to the full session—not only to the tokens beyond the threshold. (
developers.openai.com) - Regional processing: OpenAI lists a 10% uplift for GPT-5.5 regional-processing endpoints. Include it only if your setup uses one. (
developers.openai.com) - Batch or other processing modes: OpenAI’s model page notes long-context pricing across standard, batch, and flex processing, but the standard rate table above should not be treated as a complete quote for every processing mode. Select the applicable mode on the official pricing page before estimating those jobs. (
developers.openai.com) - ChatGPT subscription: ChatGPT subscription billing and API usage billing are separate. A subscription does not, by itself, mean your API usage is covered; OpenAI says API usage is billed through API billing. (
help.openai.com)
Production budget checks
Before committing production traffic, confirm:
- Provider and model: You are pricing the OpenAI API model
gpt-5.5, rather than a third-party gateway or a different model variant. - Rate tier: Your request is below or above the 272K input-token threshold, and you have applied the corresponding rate to the full session when required.
- Actual usage: Estimate input, cached input, and output separately from representative requests; output has its own, higher per-token rate.
- Other charges: Check whether your workflow uses billable tools, storage, or a regional-processing endpoint.
- Real workload: Pilot representative traffic and compare observed token use and cost with your estimate before scaling. OpenAI’s model documentation also notes that rate limits depend on usage tier, so verify the limits for your API organization if throughput affects the budget. (
developers.openai.com)
The formula is a planning estimate, not a fixed monthly price: your total depends on token volumes, context tier, model variant, processing mode, and any applicable additional charges.
Sources
Managing billing for ChatGPT and the API platform | OpenAI Help Center