Antigravity SDK adds local model support for offline agentic workflows with Gemma 4 26B
AI_Andrew · x · 2026-09-24
Google announced Antigravity SDK v0.1.18 with local model support: fully offline private agentic workflows plus hybrid cloud-local orchestration, initially optimized via Google AI Edge's LiteRT for Gemma 4 26B A4B (also works with Ollama, LM Studio, vLLM).
Key points
- Local execution benefits: no API costs or rate limits, privacy/compliance for code and requests, offline resiliency, hybrid local-cloud routing
- Default task scheduling: agents can defer workflows across conversation turns via BuiltinTools.SCHEDULE
- Custom subagent models: route specialized subagent tasks to dedicated model tiers via SubagentConfig.model
- Getting started: recommended >24GB VRAM/unified memory; pip install google-antigravity litert-lm, then import the litert-community Gemma 4 26B A4B weights from Hugging Face
More from coding & agent
- That Opus 5.5 Demoscene Demo Used Under 3% of a Weekly Usage Quota — gandamu_ml · 2026-09-24
- Christoph Nakazawa: Frameworks Still Matter in the AI Era Because Attention Is Finite — cnakazawa · 2026-09-24
- Bespoke Labs Launches AutoResearchExam: Benchmarking 24-Hour Auto-Research Agents — AlexGDimakis · 2026-09-24
- Claude Opus 5.5 One-Shots a 1990s-Style Demoscene Demo in C/C++ and OpenGL — gandamu_ml · 2026-09-24
- Claude Models, Textures and Rigs a Blender Character in Pure Code, Packaged as a Game Dev Skill — majidmanzarpour · 2026-09-24
- SIFT paper decouples tree search from costly benchmarks to make recursive self-improving coding agents scale — burkov · 2026-09-24