Bring Your Own Brain: Connecting Custom LLMs to Antigravity

Google Antigravity ships with Gemini 1.5 Pro, which is a beast of a model. But we believe in choice. Maybe you prefer the reasoning of GPT-4o, the coding style of Claude 3.5 Sonnet, or the privacy of a local Llama 3.

Antigravity is Model Agnostic. Here is how to bring your own brain.

Connecting OpenAI (GPT-4o)

1. Go to Settings > AI Models.
2. Select OpenAI.
3. Enter your API Key (from `platform.openai.com`).
4. Select your default model: `gpt-4o` or `gpt-4-turbo`.

Pro Tip: You can set GPT-4o as your “Reasoning” model while keeping Gemini Flash as your “Autocomplete” model to save money.

Connecting Anthropic (Claude 3.5 Sonnet)

Claude 3.5 Sonnet is widely considered the best coding model right now.

1. Go to Settings > AI Models.
2. Select Anthropic.
3. Enter your API Key.
4. Select `claude-3-5-sonnet`.

Connecting Local Models (Ollama)

Want to run offline?

1. Ensure Ollama is running (`ollama serve`).
2. Go to Settings > AI Models.
3. Select Local / Custom.
4. Enter the endpoint: `http://localhost:11434`.
5. Enter the model name: `deepseek-coder-v2`.

The “Router” Mode

This is the killer feature. You don't have to pick just one.
Enable Router Mode, and Antigravity will dynamically switch models based on the task complexity.

* User types: “Fix this typo.” -> Router: Uses `Gemini Flash` (Cheap/Fast).
* User types: “Refactor this entire class hierarchy.” -> Router: Uses `Claude 3.5 Sonnet` (Smart/Expensive).

Conclusion

Your IDE shouldn't lock you into one AI provider. With Antigravity, you have access to the entire intelligence of the internet (and your local machine).

At BYS Marketing, we configure our environments to use the best model for the job, optimizing for both performance and cost.

—

Need help configuring your enterprise AI stack?
Contact BYS Marketing. We build custom AI infrastructure for development teams.