GPT-5.5 API token cost
An Inworld model listing for GPT-5.5 reports these rates in USD per 1 million tokens:
| Token category | Reported rate |
|---|---|
| Input | $5.00 per 1M tokens |
| Output | $30.00 per 1M tokens |
| Cache read | $0.50 per 1M tokens |
| Cache write | Not stated; displayed as “—” |
These are rates shown on Inworld’s listing, which describes access through Inworld LLM Inference. They are not verified here as direct OpenAI API rates. The listing does not explain the cache-write dash, define which tokens qualify for cache-read pricing, or provide an effective date in the captured pricing text.
Estimate your token spend
Using the rates that apply to your chosen provider, estimate standard input and output charges as follows:
Estimated cost = (input tokens ÷ 1,000,000 × input rate) + (output tokens ÷ 1,000,000 × output rate)
For example, using the rates displayed on the Inworld listing:
- 100,000 input tokens: 100,000 ÷ 1,000,000 × $5 = $0.50
- 20,000 output tokens: 20,000 ÷ 1,000,000 × $30 = $0.60
- Estimated total: $1.10
That is an arithmetic estimate from the listed rates—not a guaranteed invoice amount or confirmation of direct OpenAI pricing. It excludes cache-read charges and any other fees or adjustments that may apply. If you want to estimate cache reads, the listing’s displayed rate would make 1 million cache-read tokens $0.50; confirm the provider’s eligibility and billing rules before using that in a budget. Do not assume a cache-write price from the “—” entry.
Check the billing route before budgeting
The Inworld page identifies the model as openai/gpt-5.5 and shows an OpenAI-SDK-compatible example using Inworld’s https://api.inworld.ai/v1/chat/completions endpoint. That is an Inworld-hosted API route, not a direct OpenAI endpoint. An SDK-compatible example does not establish which provider will bill you or what your final charge will be.
Before treating an estimate as a production budget, confirm with the provider you plan to use:
- Its current GPT-5.5 rate card and the billing route your application will call.
- Which token categories are billed, including how cached input is identified and charged.
- Whether cache writes, billing increments, rounding, or other applicable adjustments affect the estimate.
- Actual usage and charges in the provider’s usage or billing data, when available.
The available listing supports a useful planning calculation, but it does not settle direct OpenAI pricing or all provider-specific billing rules. For production budgeting, use the rate card and billing data for your actual API provider.