Google Unveils Gemini 3.5 Flash, Omni, and Antigravity Stack
Update your integration to use gemini-3.5-flash and adjust token limits to 1,048,576 input and 65,536 output.
Update your code to use gemini-3.5-flash, adjust token limits, and review pricing to avoid cost overruns.
Summary
Google announced the general‑availability release of Gemini 3.5 Flash at its I/O event, positioning it as the company’s most powerful agentic and coding model yet. The new model boasts a 1 million‑token context window, a 65 k maximum output, four distinct “thinking” levels, and the ability to preserve thoughts across turns. Benchmarks show Gemini 3.5 Flash outperforms its predecessor Gemini 3.1 Pro on Terminal‑Bench 2.1, GDPval‑AA, and MCP Atlas, and runs four times faster than comparable frontier models, reaching up to 12× speed in Antigravity. Independent tests give it a 55 score on the Intelligence Index and rank it #9 in both Text Arena and Code Arena Frontend, though it is 5.5× costlier than Gemini 3 Flash and 75 % pricier than Gemini 3.1 Pro.
The launch also introduced Gemini Omni Flash, a multimodal extension that handles text, image, video, and audio inputs, enabling on‑the‑fly video edits and generation across Gemini, Flow, Shorts, and APIs. Antigravity 2.0 expands the developer ecosystem with a desktop client, CLI, SDK, managed agents, and Gemini Spark background agents that run on dedicated Google Cloud VMs, allowing long‑running tasks even when user devices are closed. Google reports processing 3.2 quadrillion tokens per month—up 7× YoY—and that the Gemini app now serves over 900 million monthly users in 230+ countries and 70+ languages.
In consumer products, Gemini 3.5 Flash becomes the default model in AI Mode worldwide. The AI‑powered search box is redesigned to accept multimodal inputs and offer dynamic suggestions, while new search agents launch this summer for AI Pro and Ultra subscribers. Booking features let users request services and have Google call businesses on their behalf, and mini‑apps via Antigravity can generate custom visual tools and dashboards. Personal Intelligence expands to nearly 200 countries without a subscription, and AI Mode supports follow‑up questions that carry context across sessions. Gemini 3.5 Flash’s coding capabilities are already free in Search for all users this summer.
Key changes
- New model ID gemini-3.5-flash GA
- Knowledge cutoff January 2025
- Input token limit 1,048,576 and output limit 65,536
- No computer use feature
- Beta Interactions API for server‑side history
- Price tripled vs Gemini 3 Flash Preview and sextupled vs Gemini 3.1 Flash‑Lite
- Available in Gemini app, Search AI Mode, Antigravity, API, Android Studio, Enterprise Agent Platform
- Gemini 3.5 Pro announced for next month at higher price