Google Introduces Gemini 3.5 Flash at I/O 2026: A Faster and Cheaper Model for AI Agents and Coding

Details

Google announced Gemini 3.5 Flash at Google I/O 2026, the first model in the 3.5 series. It outperforms Gemini 3.1 Pro — the previous premium tier — on challenging benchmarks while sitting in the faster, cheaper Flash category. The model scores 76.2% on Terminal-Bench 2.1 (coding), 1656 Elo on GDPval-AA (agentic tasks), 83.6% on MCP Atlas (tool-use reliability), and 84.2% on CharXiv Reasoning (multimodal understanding). Output token generation runs at 4x the speed of the prior generation at less than half the cost for most workloads. Pricing is $1.50/M input tokens, $9.00/M output tokens, and $0.15/M for cached inputs. The context window is 1,048,576 tokens, with dynamic thinking enabled by default — the model auto-allocates more compute for harder problems. Supported input modalities are text, image, audio, and video; knowledge cutoff is January 2026. Google also introduced Managed Agents in the Gemini API: a single call spins up a full agent with tool use and code execution in an isolated Linux container with persistent file state across turns. The Antigravity 2.0 desktop app orchestrates parallel subagents for production pipelines, and the Antigravity CLI offers terminal-native agent creation. Enterprise adoption is already underway — Shopify is running Gemini 3.5 Flash subagents in parallel for merchant data analysis. Google is encouraging Gemini CLI users to migrate to Antigravity now.