How to properly configure Custom/OpenAI-Compatible Models in Antigravity IDE?

Google Antigravity
1

Hi everyone,

I am exploring Antigravity (AG) IDE and would like to know if it fully supports configuring and using custom AI models (similar to how Cursor or Windsurf allow users to swap out the default models for third-party providers).

Specifically, I want to use different models for the standard AI features and the AI Agent, such as DeepSeek, Qwen, Claude Opus, etc., via OpenAI-compatible endpoints, rather than being restricted to Gemini.

I tried to customize this by modifying my .antigravity/settings.json file as follows:
{
“antigravity.ai.provider”: “openai-compatible”,
“antigravity.openai-compatible.endpoint”: “https://api.deepseek.com”,
“antigravity.openai-compatible.apiKey”: “sk-xxxx”,
“antigravity.ai.model”: “deepseek-v4-flash”,
“antigravity.agent.enabled”: true,
“antigravity.agent.provider”: “openai-compatible”,
“antigravity.agent.openai-compatible.endpoint”: “https://api.deepseek.com”,
“antigravity.agent.openai-compatible.apiKey”: “sk-xxxx”,
“antigravity.agent.model”: “deepseek-v4-pro”,
“antigravity.fallback.enabled”: false,
“antigravity.fallback.chain”: [
{
“provider”: “openai-compatible”,
“model”: “deepseek-v4-pro”,
“endpoint”: “https://api.deepseek.com”,
“apiKey”: “sk-xxx”
}
]
}

The Issue: Despite applying these settings, it doesn’t seem to work. The IDE either fails to route the requests to the custom endpoint or defaults back, and I cannot find a dedicated model picker UI to dynamically switch between these custom configurations

My Questions:

  1. Does Antigravity IDE currently support the openai-compatible provider out of the box, or are these configuration keys named differently?

  2. Are there any plans to introduce a native Model Picker in the UI to make switching custom models as seamless as it is in Cursor?

Any guidance, correct schema definitions, or workarounds to get custom models running would be greatly appreciated!

Thanks in advance