Hi,
I am testing the Microsoft Foundry Responses API using the following project endpoint:
POST https://FOUNDRY_RESOURCE_NAME.services.ai.azure.com/api/projects/PROJECT_NAME/openai/v1/responses
The following request works correctly when using gpt-6-sol:
{
"model": "gpt-6-sol",
"input": "Search the web and tell me the current Microsoft Foundry homepage title.",
"tools": [
{
"type": "web_search"
}
],
"tool_choice": "required"
}
However, when I send the same request using gpt-6-luna:
{
"model": "gpt-6-luna",
"input": "Search the web and tell me the current Microsoft Foundry homepage title.",
"tools": [
{
"type": "web_search"
}
],
"tool_choice": "required"
}
I receive the following error:
{
"error": {
"message": "Tool 'web_search_preview' is not supported with gpt-6-luna-2026-09-22.",
"type": "invalid_request_error",
"param": "tools",
"code": null,
"request_id": "102ed4b84867293564bf24b783a89352"
}
}
Both deployments are in the same Microsoft Foundry resource/project, and the only change between the two requests is the model/deployment name.
My questions are:
- Is
web_search officially supported with gpt-6-luna-2026-09-22 in Microsoft Foundry?
- If not, is this a documented limitation?
- If it is supposed to be supported, is the error above a known issue?
- Why does the error refer to
web_search_preview when the request explicitly uses "type": "web_search"?
The same request works as expected with gpt-6-sol, so this appears to be specific to gpt-6-luna.
Any clarification on the intended capability matrix for GPT-6 Luna versus GPT-6 Sol would be appreciated.