How to Access ChatGPT 5.0: ChatGPT and API Options

Short answer (November 12, 2026): You can’t select GPT-5 Instant or Thinking in ChatGPT; OpenAI says it retired those models from ChatGPT on February 13, 2026. If you want a chat interface, sign in to ChatGPT and choose a model available in your account. If you want GPT-5 for an application, check your API project: OpenAI’s deprecation documentation describes a December 11, 2026 shutdown for older GPT-5 snapshots, but that notice alone does not establish the lifecycle or availability of the gpt-5 alias in your project. (developers.openai.com)

“ChatGPT 5.0” is often used as a search phrase; the retirement notice calls the ChatGPT models GPT-5 Instant and Thinking. ChatGPT and API access are separate routes. (developers.openai.com)

If you want to use ChatGPT

1. Sign in to ChatGPT.
2. Open the model menu and choose an option shown for your account.
3. If you don’t see an expected option, check that you’re using the right account or workspace. Available models can depend on your plan and workspace settings. (developers.openai.com)

GPT-5 Instant and Thinking are retired in ChatGPT, so looking for them in the model menu won’t restore access. OpenAI’s ChatGPT documentation describes newer model options; check what is available to your account. If an older guide says GPT-5 was the default or tells you to choose “GPT-5 Thinking,” those are historical launch instructions, not current steps. (developers.openai.com)

If you mean the OpenAI API

API availability has a separate lifecycle. OpenAI’s deprecation guidance says to check the model list’s optional shutdown_date field for a particular model. The supplied notice about older GPT-5 snapshots does not, by itself, tell you whether the gpt-5 alias is available to your project or whether it shares those snapshots’ shutdown date. A null shutdown date means no shutdown date has been announced for that listed model; it is not a promise of indefinite access. (developers.openai.com)

1. Set up API access

1. Sign in to the OpenAI API Platform and use a project where you have permission to make API requests.
2. Create a project API key and store it securely. For a deployed application, use a project service account and its key rather than sharing a developer’s personal key. (developers.openai.com)
3. Set up API billing if needed for paid usage. API billing and charges are separate from a ChatGPT subscription; API organization owners can manage billing for their organization. (help.openai.com)
4. Set the key in your terminal as OPENAI_API_KEY. OpenAI’s SDKs read the key from the environment. (developers.openai.com)

For macOS or Linux:

export OPENAI_API_KEY="your_api_key_here"

2. Check your project’s model list

Run this in a terminal with curl and jq installed:

curl -sS https://api.openai.com/v1/models \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  | jq '.data[] | select(.id == "gpt-5") | {id, shutdown_date}'

OpenAI documents GET /v1/models as the model-list endpoint and says to inspect each model’s optional shutdown_date. If this command prints no result for gpt-5, the model is not in the list returned to that project. If it prints an entry, check its shutdown_date and confirm the model works with a request before relying on it. The public notice retrieved for this guide does not settle whether the alias shares the older snapshots’ December 11, 2026 shutdown. (developers.openai.com)

3. Make a first request

Install the official Python SDK:

pip install openai

Save this as example.py and run python example.py. The example uses gpt-6-astra, the model identifier in OpenAI’s quickstart; it is a current-model example, not GPT-5. If your model-list check confirms gpt-5 is available to your project, you can substitute gpt-5 as the model value to test it. (developers.openai.com)

from openai import OpenAI

client = OpenAI()
response = client.responses.create(
    model="gpt-6-astra",
    input="Write a one-sentence bedtime story about a unicorn.",
)
print(response.output_text)

OpenAI recommends setting up API access and making a first request through its developer quickstart. A successful request confirms that the chosen model works for that project at that time; check its lifecycle before depending on it in production. (developers.openai.com)

Which route should you use?

• You want to chat: Use ChatGPT and select a model shown for your account. GPT-5 Instant and Thinking are retired there.
• You’re building an application: Use an API project and a project key. Check the project’s model list and lifecycle details, then test the selected model before deployment.
• You need GPT-5 specifically: ChatGPT is not the route. The API model-list check can show whether gpt-5 is listed for your project, but the public notice available here does not establish the alias’s exact shutdown condition. Do not treat a missing shutdown_date as guaranteed ongoing access. (developers.openai.com)

Sources

Retiring GPT-4o and other ChatGPT models — OpenAI Help Center

GPT-6 and other models in ChatGPT — OpenAI Help Center

Deprecations — OpenAI API

Developer quickstart — OpenAI API

API keys and service accounts — OpenAI API

List models — OpenAI API Reference

API Overview — OpenAI API Reference

Managing billing for ChatGPT and the API platform — OpenAI Help Center