Codex can use a user-level custom provider when a team wants one OpenAI-compatible gateway for a coding client and its own applications. This guide focuses on configuration checks before production use.
Choose a GPT-5.6 model
The live ApiHub catalog may list several GPT-5.6 variants. Confirm the exact model ID and availability in the catalog before running a request.
- Use a higher-capability variant for complex, open-ended coding work.
- Use a balanced variant for everyday development and debugging.
- Use a lightweight variant for clear, repeatable batch tasks.
These are workflow labels, not a promise of model behavior.
Configure the provider
Create an API key in the ApiHub console, then set it in your shell:
export APIHUB_API_KEY=your_api_key
On Windows PowerShell:
$env:APIHUB_API_KEY=your_api_key
Add the provider to your user-level config.toml:
model = gpt-5.6-terra
model_provider = apihub
[model_providers.apihub]
name = ApiHub
base_url = https://video.gqaiapp.com/v1
env_key = APIHUB_API_KEY
wire_api = responses
Use a model ID currently shown in the live catalog. Check whether your Codex version expects user-level configuration.
Verify before production use
- Confirm the Base URL and API key scope.
- Run a non-sensitive test request.
- Check the response format and usage record.
- Confirm how failures, refunds, data retention, and upstream routing are handled.
- Do not send secrets, personal data, or confidential production code until your own security and compliance review is complete.
ApiHub is an independent third-party service and is not affiliated with OpenAI or Anthropic. Model availability and pricing can change; the live catalog is the source of truth.
I am interested in feedback on Codex provider configuration, Responses compatibility, and documentation gaps.
