Yes. GPT-5.5 is available through OpenAI’s direct API. Use gpt-5.5 as the model ID. OpenAI lists both the Responses API and Chat Completions as supported endpoints; this quickstart uses Responses. OpenAI announced API availability on April 24, 2026. (OpenAI)
Make your first GPT-5.5 API request
You need an OpenAI API key for your project. The example below uses the official Python SDK and Responses API. A key by itself does not guarantee that a particular project can use a model, so confirm model access and billing in your API account if a request is rejected. API use is billed separately according to usage. (OpenAI API quickstart)
Install the SDK and set the key in your shell environment:
pip install openai
export OPENAI_API_KEY="your-api-key"
Save this as example.py and run it with Python:
from openai import OpenAI
client = OpenAI()
response = client.responses.create(
model="gpt-5.5",
input="In one sentence, explain what an API does.",
)
print(response.output_text)
The SDK’s default client reads OPENAI_API_KEY from the environment. The request sends model and input to the Responses API; response.output_text prints the generated text. (OpenAI API quickstart)
The model page lists gpt-5.5 as an alias and gpt-5.5-2026-04-23 as a snapshot for pinning a specific version. Use the alias for the standard quickstart above, or the snapshot when you need to target that version explicitly. (GPT-5.5 model page)
GPT-5.5 API capabilities and limits
OpenAI lists a 1,050,000-token context window and a 128,000-token maximum output. The model accepts text and image input and returns text; audio and video are listed as unsupported. Streaming, function calling, and structured outputs are supported. (GPT-5.5 model page)
If you plan to use tools with the Responses API, check the current model documentation for compatibility and the relevant tool documentation for setup. Do not assume a tool is available or configured just because the API supports tool integrations.
The documented reasoning-effort options are none, low, medium, high, and xhigh; medium is the default. OpenAI recommends treating GPT-5.5 as a model to tune for, not a drop-in replacement: establish a fresh baseline and evaluate prompts and settings against representative examples. (GPT-5.5 model page; Using GPT-5.5)
GPT-5.5 API pricing
OpenAI lists these prices in U.S. dollars per 1 million tokens: (GPT-5.5 model page)
| Token category | Price |
|---|---|
| Input | $5.00 / 1M tokens |
| Cached input | $0.50 / 1M tokens |
| Output | $30.00 / 1M tokens |
There are pricing conditions to account for. For prompts with more than 272,000 input tokens, input is priced at 2× and output at 1.5× for the full session. OpenAI also lists a 10% uplift for regional-processing (data-residency) endpoints. (GPT-5.5 model page)
Example estimate: 10,000 uncached input tokens cost $0.05 at the listed rate; 2,000 output tokens cost $0.06. That is $0.11 total, assuming the request is below the long-context threshold and has no other applicable endpoint conditions. (GPT-5.5 model page)
Before putting it in production
- Confirm your project can access
gpt-5.5and review your project’s current API limits. - Test representative prompts, output formats, and any tools you plan to use. OpenAI recommends establishing a fresh baseline when migrating from an earlier model. (Using GPT-5.5)
- Estimate costs using input, cached-input, and output tokens, including the long-context pricing condition when it applies. (GPT-5.5 model page)
- Keep provider identifiers distinct.
gpt-5.5is the documented direct OpenAI API model ID; a gateway’s route name is not automatically interchangeable.
Direct API access is different from using GPT-5.5 in ChatGPT or Codex, or calling it through a third-party gateway. Those surfaces can have different access, billing, and integration details. (OpenAI)