GPT-5.1-chat API Parameters Fix

Date: December 31, 2025
Issue: Chat endpoint failing with LLM API errors
Root Cause: gpt-5.1-chat model has different API parameter requirements

Problem

Chat endpoint was returning error messages because the LLM API calls were failing with:

  • 400 Bad Request: Unsupported parameter: 'max_tokens' is not supported with this model. Use 'max_completion_tokens' instead.
  • 400 Bad Request: Unsupported value: 'temperature' does not support 0.7 with this model. Only the default (1) value is supported.

Root Cause

The gpt-5.1-chat model (version 2025-11-13) has different API parameter requirements than older models:

  1. max_tokens → max_completion_tokens: The model requires max_completion_tokens instead of max_tokens
  2. Temperature not supported: The model only supports the default temperature value (1), custom values like 0.7 are not allowed
  3. API version: Requires 2024-12-01-preview or later (not 2024-05-01-preview)

Solution

Updated backend/agents/base.py FoundryChatClient class to:

  1. Use max_completion_tokens for Azure format endpoints (instead of max_tokens)
  2. Skip temperature parameter for gpt-5.1-chat models (only send for older models)
  3. Store deployment name to enable model-specific handling

Code Changes 

# Store deployment name for model-specific handling
self.deployment = deployment

# In ainvoke method:
if not self.is_openai_compat:
    # gpt-5.1-chat doesn't support custom temperature, so don't send it
    # For older models, temperature is supported
    if self.deployment and "gpt-5.1" not in self.deployment.lower():
        payload["temperature"] = self.temperature
    # Use max_completion_tokens for newer models (gpt-5.1-chat, etc.)
    payload["max_completion_tokens"] = self.max_tokens

Testing

Before Fix 

# Test with max_tokens and temperature
curl -X POST \
  -H "api-key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Say hi"}],
    "temperature": 0.7,
    "max_tokens": 100
  }' \
  "https://zimax-gw.azure-api.net/zimax/openai/deployments/gpt-5.1-chat/chat/completions?api-version=2024-12-01-preview"

# Result: 400 Bad Request

After Fix 

# Test with max_completion_tokens and no temperature
curl -X POST \
  -H "api-key: $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [{"role": "user", "content": "Say hi"}],
    "max_completion_tokens": 100
  }' \
  "https://zimax-gw.azure-api.net/zimax/openai/deployments/gpt-5.1-chat/chat/completions?api-version=2024-12-01-preview"

# Result: 200 OK ✅

Configuration Requirements

Ensure these environment variables are set correctly:

  • AZURE_AI_ENDPOINT = https://zimax-gw.azure-api.net/zimax (base URL, no /openai/v1/)
  • AZURE_AI_DEPLOYMENT = gpt-5.1-chat
  • AZURE_AI_API_VERSION = 2024-12-01-preview (required for gpt-5.1-chat)
  • AZURE_AI_MODEL_ROUTER = (empty or not set)
  • AZURE_AI_KEY = (API key from Key Vault)

Model Compatibility

Model max_tokens max_completion_tokens temperature API Version
gpt-4, gpt-35-turbo ✅ ❌ ✅ (custom) 2024-05-01-preview
gpt-5.1-chat ❌ ✅ ❌ (default only) 2024-12-01-preview
  • docs/troubleshooting/api-version-model-version-mismatch.md - API version issues
  • docs/troubleshooting/chat-error-diagnosis.md - General chat troubleshooting
  • docs/configuration/config-alignment.md - Configuration reference