Skip to content

Why can't I access the 1M-token model context window when using the Azure API? #34856

Description

@ZhengXiaoen

What version of the Codex App are you using (From “About Codex” dialog)?

26.715.12143

What subscription do you have?

Azure PayGo

What platform is your computer?

Windows 11

What issue are you seeing?

The model’s context length should not be entirely controlled by the client. I use Codex because it’s a good product, but the backend I’m connected to is Azure OpenAI, which already supports a 1M-token context window. The client should not hard-code the context length and impose restrictions on my usage. I understand that requests exceeding 272K tokens may incur additional charges, but since I’m using a Pay-As-You-Go subscription, users like us should not be subject to such limitations.

What steps can reproduce the bug?

none

What is the expected behavior?

No response

Additional information

No response

Metadata

Metadata

Assignees

No one assigned

    Labels

    appIssues related to the Codex desktop appazureIssues related to the Azure-hosted OpenAI modelscontextIssues related to context management (including compaction)enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions