Short answer
You can’t select the original GPT-5 Instant or Thinking in ChatGPT: OpenAI says those models were retired there on February 13, 2026. GPT-5 is still listed in OpenAI’s API documentation as deprecated, with the alias gpt-5, the snapshot gpt-5-2025-08-07, and the Responses endpoint. Treat it as a legacy API option, not a model to build a new production app around. (OpenAI Help Center · GPT-5 API model page)
- Want to chat? Sign in to ChatGPT and choose among the models currently shown in its model menu. The original GPT-5 Instant and Thinking are retired there; a paid plan does not restore them.
- Need GPT-5 in software? The API model page marks GPT-5 deprecated. Check that page and your project’s access before relying on it.
Can I still get GPT-5 in ChatGPT?
No. OpenAI says GPT-5 Instant and Thinking are no longer available in ChatGPT. If you don’t see GPT-5 in the model menu, that is expected—not a sign-in, app-update, or subscription problem. Older conversations may be directed to newer model equivalents; that does not mean those conversations are still using GPT-5. (OpenAI Help Center)
To use ChatGPT, sign in and choose an option currently shown in the model menu. Available models and controls can depend on your plan and workspace. Don’t assume similarly named models are interchangeable; check the name shown in ChatGPT. (OpenAI Help Center)
How to try GPT-5 through the API
OpenAI’s API model page marks GPT-5 deprecated and lists the alias gpt-5 and snapshot gpt-5-2025-08-07. It lists the Responses and Chat Completions endpoints. Check the current model documentation and confirm your project can access GPT-5 before integrating it; deprecated status does not guarantee future availability. (GPT-5 API model page)
1. Set up API access
- Sign in to the OpenAI API Platform and create an API key.
- Set up API billing separately from ChatGPT. The two products use separate billing systems; paid API usage requires API billing setup. (OpenAI Help Center)
- Install the Python SDK and make your API key available to your environment as
OPENAI_API_KEY. The official Python library documentspip install openaiand reads that environment variable when you initialize the client. (OpenAI Python library)
pip install openai
For a production application, keep credentials on your server or in a secure secret store. Don’t put an API key in browser or mobile-app code or commit it to a repository. (API key safety guidance)
2. Make a minimal request
Save this as try_gpt5.py and run python try_gpt5.py after setting OPENAI_API_KEY in your environment:
from openai import OpenAI
client = OpenAI()
response = client.responses.create(
model="gpt-5",
input="Reply with one sentence explaining what an API is."
)
print(response.output_text)
print("Returned model:", response.model)
The official Python library uses this Responses API pattern and exposes generated text through response.output_text. The returned response.model value is useful for inspecting the model identifier in the response; don’t treat it as proof that an alias resolved to a particular dated snapshot. (OpenAI Python library · Python Response type)
3. Read the listed price correctly
The GPT-5 model page labels its displayed token rates as Batch API prices, not standard real-time rates: $1.25 per 1 million input tokens, $0.125 per 1 million cached input tokens, and $10 per 1 million output tokens. These are USD rates per million tokens for Batch API usage; don’t use them as a real-time price estimate. Recheck current pricing before budgeting. (GPT-5 API model page)
For illustration only, 100,000 uncached input tokens and 20,000 output tokens at those listed Batch rates would cost about $0.325: (100,000 ÷ 1,000,000 × $1.25) + (20,000 ÷ 1,000,000 × $10). This is a Batch-rate calculation, not a quote for a real-time request. It excludes any separate charges and assumes those listed rates apply to the usage. API billing is separate from a ChatGPT plan. (GPT-5 API model page · OpenAI Help Center)
Is GPT-5 a good choice for a new production app?
Usually, don’t start a new production integration with a model OpenAI marks deprecated unless you have a specific compatibility need and have confirmed it still works for your project. The GPT-5 page recommends using a newer model. For new production work, select a currently supported model and test it against your requirements before migrating. (GPT-5 API model page)
If you maintain a legacy GPT-5 integration, monitor API usage and errors and keep a migration plan. If a request is rejected, check the current model page, your project’s access, API billing, and applicable usage limits. (GPT-5 API model page)
Bottom line: The original GPT-5 Instant and Thinking are no longer available in ChatGPT. GPT-5 remains listed as a deprecated API model, but check current documentation and project access before using it. For a new production application, choose a currently supported model.
Sources
Retiring GPT-4o and other ChatGPT models
Python Responses API response type
Managing billing for ChatGPT and the API platform