Google Antigravity SDK Adds Local Model Support for Greater Token Efficiency and Code Privacy
Google’s Antigravity SDK can split agent work between a cloud planner and local coding agents, keeping source code and implementation tasks on the developer’s machine while using cloud models for higher-level planning.
Google’s demo reports 95 of roughly 3,400 tokens sent to the cloud, with local models handling the remaining 97.2%; that token share alone does not establish lower end-to-end cost, faster runs, or better results. The SDK initially targets Gemma through LiteRT and also supports OpenAI-compatible local servers such as Ollama, LM Studio, and vLLM. InfoQ relays mixed community reactions: interest in privacy and model control, alongside calls to measure patch acceptance, latency, and hardware cost.