---
title: "Introducing Support for Local AI Models in the Antigravity SDK"
description: "The Antigravity SDK now supports local AI models, starting with Gemma 4 26B A4B via Google AI Edge’s LiteRT, enabling offline agentic assistance on local GPU and RAM with a recommended machine of 24GB VRAM or unified memory. Local execution eliminates API costs and rate limits while keeping code and requests on-device, offering privacy and offline resiliency. The SDK supports hybrid workflows where a cloud model like Gemini 3.8 Flash plans while local Gemma 4 instances execute, as in a code auditing task where 97.2% of tokens (3,322 of 3,417) ran locally. A built-in example generates a live-updating CLI resource monitor using psutil and rich from a single prompt, with the agent autonomously writing the script, creating requirements.txt, and testing the code entirely offline. The SDK also supports any OpenAI-compatible server like Ollama, LM Studio, or vLLM via LocalOpenAIAgentConfig without changing agent orchestration."
---

# Introducing Support for Local AI Models in the Antigravity SDK

[Google](https://yomu.fyi/company/google) · Sachin Kotwani, Taylor Mullen · Sep 23, 2026

**Type:** Announcement

## Summary

The Antigravity SDK now supports local AI models, starting with Gemma 4 26B A4B via Google AI Edge’s LiteRT, enabling offline agentic assistance on local GPU and RAM with a recommended machine of 24GB VRAM or unified memory. Local execution eliminates API costs and rate limits while keeping code and requests on-device, offering privacy and offline resiliency. The SDK supports hybrid workflows where a cloud model like Gemini 3.8 Flash plans while local Gemma 4 instances execute, as in a code auditing task where 97.2% of tokens (3,322 of 3,417) ran locally. A built-in example generates a live-updating CLI resource monitor using psutil and rich from a single prompt, with the agent autonomously writing the script, creating requirements.txt, and testing the code entirely offline. The SDK also supports any OpenAI-compatible server like Ollama, LM Studio, or vLLM via LocalOpenAIAgentConfig without changing agent orchestration.

## Context

Google announces local AI model support in the Antigravity SDK, enabling offline agentic workflows with cost, privacy, and resiliency benefits.

## Approach / What changed

The Antigravity SDK adds local model support via LiteRT for Gemma 4 26B A4B, using LiteRTAgentConfig and LocalOpenAIAgentConfig for OpenAI-compatible backends, with optimized local GPU and RAM usage.

## Takeaways

- Local model execution in the Antigravity SDK eliminates API costs and rate limits, and keeps all data on-device for privacy and offline use.
- Hybrid workflows can combine a cloud architect model like Gemini 3.8 Flash with local Gemma 4 instances, achieving 97.2% local token utilization.
- The SDK supports any OpenAI-compatible backend (Ollama, LM Studio, vLLM) via LocalOpenAIAgentConfig without altering agent orchestration.

## FAQ

### What hardware is recommended for local models?

A machine with more than 24GB VRAM or unified memory is recommended for running Gemma 4 26B A4B locally.

### Can I use a different local inference server?

Yes, the SDK supports any OpenAI-compatible server via LocalOpenAIAgentConfig, such as Ollama, LM Studio, or vLLM, without changing agent orchestration.

### What token savings were demonstrated?

In the hybrid auditing example, 97.2% of tokens (3,322 out of 3,417) ran locally, with only 95 cloud tokens used.

**Tags:** [AI Agents](https://yomu.fyi/topic/ai-agents), [Antigravity SDK](https://yomu.fyi/topic/antigravity), [Gemma](https://yomu.fyi/topic/gemma), [LiteRT-LM](https://yomu.fyi/topic/litert-lm), [LLMs](https://yomu.fyi/topic/llm)

- Source: [Google](https://developers.googleblog.com/introducing-support-for-local-ai-models-in-the-antigravity-sdk)
- Source URL: https://developers.googleblog.com/introducing-support-for-local-ai-models-in-the-antigravity-sdk
- Ingested by Yomu: 2026-09-23T18:00:11.606Z

[Read original post](https://developers.googleblog.com/introducing-support-for-local-ai-models-in-the-antigravity-sdk)