Currently, Cursor does not provide full configuration support for reasoning models (e.g., GPT-5 and other Azure reasoning-enabled models). While basic integration is possible, several important customization options are missing.
I’d like to request the following features:
1. Reasoning Control
-
Ability to set reasoning levels: Low, Medium, High
-
Option to adjust reasoning verbosity (concise vs. detailed explanations)
2. Multiple Deployment Profiles
-
Ability to configure and save multiple model deployments with different settings (e.g., Low reasoning vs. High reasoning)
-
Quick switching between profiles for different workflows
3. Token Configuration
-
Option to define input and output token limits per deployment
-
Useful for balancing cost, performance, and response length
4. Reasoning Trace Visibility
-
Option to view reasoning traces (where supported by Azure)
-
Improves debugging, evaluation, and prompt optimization
5. Request Caching for Custom API Keys
-
Ability to cache API requests by default when using custom Azure API keys
-
Reduces latency, improves performance, and avoids redundant calls for repeated queries
At present, Cursor does not expose these options, even though they are supported in Azure AI Foundry. Enabling them would unlock the full potential of GPT-5 and other reasoning models inside Cursor.