Switched to Antigravity 6 months ago—a few thoughts on the 2.0 update and my current workflow
I moved from VSCode to Antigravity about 6 months ago and haven’t looked back. It’s been a solid driver for my development across TypeScript, JavaScript, Python, and some fairly complex HTML. The latest IDE update has been working well.
I was originally holding off on Antigravity 2.0 because I have other tools for "vibe coding" and non-coding tasks, but I finally installed it last week and I'm really impressed. It has effectively become my primary interface for almost everything else now.
A few notes on how I’m running it:
*Orchestration: I’ve configured both 2.0 and the IDE to use Gemini for local LLM orchestration. It handles my local LLM (Gemma-4 31b mostly) surprisingly well when I offload specific tasks to it.
*Model Selection: Gemini 3.5 medium is my new workhorse for code, moderate architecture, planning and spec writing. I only bump up to "high" if I hit a complex blocker or to review architecture details and plans. I've also been happy with how capable 3.1 low and flash are —they’re my go-to for light-duty tasks to keep token usage minimal and fast.
*Tool Calling: I haven't pushed this with a massive number of external tools like some of you, but for my current stack, it’s been rock solid.
For anyone on the fence or just starting out with the 2.0 update, it’s definitely worth trying out.
I’m curious—how are others handling their model-switching workflow between the 3.5 series and smaller models? Also, for those of you running local models alongside Gemini, what’s your strategy for balancing the two? I’m currently leaning on Gemma-4 31b for specific tasks (including coding), but I’d love to hear how you’re optimizing your local-to-cloud hand-offs.