404 Resource not found when calling GPT-5-Chat via Azure AI Foundry inference endpoint

Question

404 Resource not found when calling GPT-5-Chat via Azure AI Foundry inference endpoint

sagul jafarullah 0

What I’ve tried:

Verified the endpoint is correct and ends with /v1.

Confirmed the key is from the Foundry inference deployment (not Azure OpenAI).

Tried both the OpenAI Python SDK and cURL.

Ensured the model name in the request is exactly gpt-5-chat.

Tested immediately after deployment and waited >10 minutes in case of propagation delay.

Attempted both Authorization: Bearer <key> and api-key: <key> headers.

Sample request (PowerShell):

powershell
Copy
$endpoint = "https://sagul-me43r8sb-swedencentral.services.ai.azure.com/v1/chat/completions"
$key      = "<FOUNDRY_INFERENCE_KEY>"

$body = @{
  model    = "gpt-5-chat"
  messages = @(
    @{ role = "system"; content = "You are helpful." },
    @{ role = "user";   content = "Say hello" }
  )
  max_tokens = 50
} | ConvertTo-Json -Depth 5

Invoke-RestMethod -Method Post -Uri $endpoint `
  -Headers @{ "api-key" = $key; "Content-Type" = "application/json" } `
  -Body $body

Error returned:

json
Copy
{"error":{"code":"404","message":"Resource not found"}}

I need guidance on:

Whether api-key authentication is supported for Foundry inference endpoints.

If there’s a configuration step required in Azure AI Foundry for the endpoint to be usable with the SDK.

Any known issues with gpt-5-chat in the Sweden Central region.What I’ve tried:
- Verified the endpoint is correct and ends with /v1.
- Confirmed the key is from the Foundry inference deployment (not Azure OpenAI).
- Tried both the OpenAI Python SDK and cURL.
- Ensured the model name in the request is exactly gpt-5-chat.
- Tested immediately after deployment and waited >10 minutes in case of propagation delay.
- Attempted both Authorization: Bearer <key> and api-key: <key> headers.
Sample request (PowerShell):
```
  powershell
  Copy
  $endpoint = "https://sagul-me43r8sb-swedencentral.services.ai.azure.com/v1/chat/completions"
```

$key = "<FOUNDRY_INFERENCE_KEY>"

$body = @{ model = "gpt-5-chat" messages = @( @{ role = "system"; content = "You are helpful." }, @{ role = "user"; content = "Say hello" } ) max_tokens = 50 } | ConvertTo-Json -Depth 5

Invoke-RestMethod -Method Post -Uri $endpoint -Headers @{ "api-key" = $key; "Content-Type" = "application/json" } -Body $body

  
  **Error returned:**
  
  ```yaml
  json
  Copy
  {"error":{"code":"404","message":"Resource not found"}}

I need guidance on:

Whether api-key authentication is supported for Foundry inference endpoints.
If there’s a configuration step required in Azure AI Foundry for the endpoint to be usable with the SDK.
Any known issues with gpt-5-chat in the Sweden Central region.

Sina Salam 22,576 Reputation points Volunteer Moderator

2025-08-10T15:57:41.5966667+00:00
Hello sagul jafarullah,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you are having 404 Resource not found error, when calling GPT-5-Chat via Azure AI Foundry inference endpoint.

Multiple users reported similar issues with GPT-5 deployments timing out or returning 404 errors, even in other regions like East US. Either internal or not. The error message 404 Resource not found typically means the endpoint or model is not accessible, not just misconfigured. This could be due to:

Model not deployed correctly

Region not supported

Incorrect endpoint path or headers

Therefore, to resolve the issues, try the followings:

Use Azure CLI or Portal to confirm GPT-5 availability in Sweden Central:
az cognitiveservices account list-models --name <your-resource-name> --resource-group <your-rg>
Or switch to a supported region like West Europe or Central US.

Ensure the model is fully deployed and active in Azure AI Foundry. Wait at least 30 minutes post-deployment and check the status in the Foundry dashboard.

Use Bearer token instead of api-key:
-Headers @{ "Authorization" = "Bearer $key"; "Content-Type" = "application/json" }

api-key may not be supported for Foundry inference endpoints.

Ensure the endpoint follows this format: https://<deployment-name>.<region>.services.ai.azure.com/v1/chat/completions Avoid trailing slashes or incorrect paths.

If Playground times out (even in East US), try using cURL or SDK directly. This bypasses UI issues.

I hope this is helpful! Do not hesitate to let me know if you have any other questions or clarifications.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Manas Mohanty 8,150 Microsoft External Staff Moderator

Hi sagul jafarullah

Here is the analysis based on your inputs.

Region Availability

GPT-5 is currently supported for global standard deployment for East US 2 and Sweden central Global Standard and Data zone deployment.

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/reasoning?tabs=gpt-5%2Cpython-secure%2Cpy#availability

Endpoint format

The endpoint used in your code is using v1/chat completion and not using api version too

  $endpoint = "https://sagul-me43r8sb-swedencentral.services.ai.azure.com/v1/chat/completions"

As per documentation

Suggested syntax for text chat completion is as below

POST https://{resource}.services.ai.azure.com/models/chat/completions?api-version=2024-05-01-preview
{
  "modalities": [
    "text"
  ],
  "messages": [
    {
      "role": "system",
      "content": "You are a helpful assistant"
    },
    
     "role": "user", 
     "content": "Explain Riemann's conjecture" 
    }
  "frequency_penalty": 0,
  "presence_penalty": 0,
  "temperature": 0,
  "top_p": 0,
  "seed": 21,
  "model": "my-model-name"
}]

Correct Usage in Azure OpenAI SDK


import os
import base64
from openai import AzureOpenAI

endpoint = os.getenv("ENDPOINT_URL", "https://<yourazureopaniname>.openai.azure.com/") deployment = os.getenv("DEPLOYMENT_NAME", "gpt-5")
subscription_key = os.getenv("AZURE_OPENAI_API_KEY", "REPLACE_WITH_YOUR_KEY_VALUE_HERE")

# Initialize Azure OpenAI client with key-based authentication
client = AzureOpenAI(
    azure_endpoint=endpoint,
    api_key=subscription_key,
    api_version="2025-01-01-preview",
)

# IMAGE_PATH = "YOUR_IMAGE_PATH"
# encoded_image = base64.b64encode(open(IMAGE_PATH, 'rb').read()).decode('ascii')

# Prepare the chat prompt
chat_prompt = [
    {
        "role": "developer",
        "content": [
            {
                "type": "text",
                "text": "You are an AI assistant that helps people find information."
            }
        ]
    },
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": "How would earth look to aliens."
            }
        ]
    }
]

# Include speech result if speech is enabled
messages = chat_prompt

# Generate the completion
completion = client.chat.completions.create(
    model=deployment,
    messages=messages,
    max_completion_tokens=16384,
    stop=None,
    stream=False
)

print(completion.to_json())

Ask

Could you also confirm that

You had O3 access prior or GPT 5 image access through Gated form already.

Thank you

Your answer

Sina Salam 22,576 Reputation points Volunteer Moderator

2025-08-10T15:57:41.5966667+00:00

Hello sagul jafarullah,

Welcome to the Microsoft Q&A and thank you for posting your questions here.

I understand that you are having 404 Resource not found error, when calling GPT-5-Chat via Azure AI Foundry inference endpoint.

Multiple users reported similar issues with GPT-5 deployments timing out or returning 404 errors, even in other regions like East US. Either internal or not. The error message 404 Resource not found typically means the endpoint or model is not accessible, not just misconfigured. This could be due to:

Model not deployed correctly

Region not supported

Incorrect endpoint path or headers

Therefore, to resolve the issues, try the followings:

Use Azure CLI or Portal to confirm GPT-5 availability in Sweden Central:
az cognitiveservices account list-models --name <your-resource-name> --resource-group <your-rg>
Or switch to a supported region like West Europe or Central US.

Ensure the model is fully deployed and active in Azure AI Foundry. Wait at least 30 minutes post-deployment and check the status in the Foundry dashboard.

Use Bearer token instead of api-key:
-Headers @{ "Authorization" = "Bearer $key"; "Content-Type" = "application/json" }

api-key may not be supported for Foundry inference endpoints.

Ensure the endpoint follows this format: https://<deployment-name>.<region>.services.ai.azure.com/v1/chat/completions Avoid trailing slashes or incorrect paths.

If Playground times out (even in East US), try using cURL or SDK directly. This bypasses UI issues.

I hope this is helpful! Do not hesitate to let me know if you have any other questions or clarifications.

Please don't forget to close up the thread here by upvoting and accept it as an answer if it is helpful.

Share via

404 Resource not found when calling GPT-5-Chat via Azure AI Foundry inference endpoint

Your answer