Copilot / gpt-5-codex "model gpt-5-codex is not accessible via the /chat/completions endpoint"

[andrewneilson]

AI_APICallError: model gpt-5-codex is not accessible via the /chat/completions endpoint

gpt-5-codex from copilot was added about 8 hours ago sst/models.dev#222

But trying to use it in opencode 0.11.3 I get the above error

I do see it available in my org

When I check the available models from the copilot api it doesn't look available, but the fact that it was added makes me wonder if I'm missing something:

gpt-4.1
gpt-5-mini
gpt-5
gpt-3.5-turbo
gpt-3.5-turbo-0613
gpt-4o-mini
gpt-4o-mini-2024-07-18
gpt-4
gpt-4-0613
gpt-4-0125-preview
gpt-4o
gpt-4o-2024-11-20
gpt-4o-2024-05-13
gpt-4-o-preview
gpt-4o-2024-08-06
o3-mini
o3-mini-2025-01-31
o3-mini-paygo
text-embedding-ada-002
text-embedding-3-small
text-embedding-3-small-inference
claude-3.5-sonnet
claude-3.7-sonnet
claude-3.7-sonnet-thought
claude-sonnet-4
claude-opus-4
claude-opus-41
gemini-2.0-flash-001
gemini-2.5-pro
o3
o3-2025-04-16
o4-mini
o4-mini-2025-04-16
gpt-4.1-2025-04-14

[OpeOginni]

Hello @andrewneilson noticed that in this url https://github.com/settings/copilot/features, you can enable different models to use for the copilot API, noted gpt-5-codex is not there to enable or disable so checked the docs, and that shows that 5-codex is only available on VS Code for now, so maybe it doesn't work on the API yet docs

[beratcmn]

Even tho I enabled the codex in settings I still have the same problem as the OG. Here is the screenshot:

[OpeOginni]

@beratcmn still shows only for VSCode in the docs

[caozhiyuan]

gpt-5-codex use openai response api https://platform.openai.com/docs/api-reference/responses/create endpoint , diff with /chat/completions , should use /responses endpoint , the request parameters have changed @OpeOginni

[OpeOginni]

@caozhiyuan thanks and yeah that makes sense, so you think its an issue on Copliots side? with how they are serving the model?

[caozhiyuan]

@OpeOginni Because OpenAI will use the responses API for its subsequent new models, Copilot also provides a responses API. this is copilot gpt-5-codex model info {
"capabilities": {
"family": "gpt-5-codex",
"limits": {
"max_context_window_tokens": 200000,
"max_output_tokens": 64000,
"max_prompt_tokens": 128000,
"vision": {
"max_prompt_image_size": 3145728,
"max_prompt_images": 1,
"supported_media_types": ["image/jpeg", "image/png", "image/webp", "image/gif"]
}
},
"object": "model_capabilities",
"supports": {
"parallel_tool_calls": true,
"streaming": true,
"structured_outputs": true,
"tool_calls": true,
"vision": true
},
"tokenizer": "o200k_base",
"type": "chat"
},
"id": "gpt-5-codex",
"model_picker_category": "powerful",
"model_picker_enabled": true,
"name": "GPT-5-Codex (Preview)",
"object": "model",
"policy": {
"state": "enabled",
"terms": "Enable access to the latest GPT-5-Codex model from OpenAI. Learn more about how GitHub Copilot serves GPT-5-Codex."
},
"preview": true,
"supported_endpoints": ["/responses"],
"vendor": "OpenAI",
"version": "gpt-5-codex"
}

[varo6]

still having the same problem, but on copilot docs they claim it is only supported in Visual Studio Code v1.104.1 or higher

[OpeOginni]

Hey did some digging and this is what I am able to find.

1. On models.dev the github-api provided models make use of the @ai-sdk/openai-compatible package.
2. as noted by @caozhiyuan gpt-5-codex only uses the /responses path
3. In the codebase for the @ai-sdk/openai-compatible package /chat/completions is hardcoded as default path

So this should be a fix on the ai-sdk side or this specific model provided on the github-api should use the @ai-sdk/openai package which supports the /responses endpoint

[melMass]

FYI the error is in vscode insiders (latest) too