Google Ships Gemini 3.5 Flash: Agentic-First Model Beats GPT-5.5 on Key Benchmarks

Google DeepMind·high signal

Google released Gemini 3.5 Flash at I/O 2026 on May 19, a model purpose-built for agentic workflows scoring 76.2% on Terminal-Bench 2.1 and leading the field on MCP Atlas (83.6%) and Toolathlon, surpassing Claude Opus 4.7 and GPT-5.5 on agentic benchmarks while outputting tokens 4x faster than other frontier models. Priced at $1.50/1M input tokens with a 1M-token context window, it's immediately available across Gemini API, Vertex AI, and Google AI Studio — though it loses to competitors on ARC-AGI-2 and 128k-token retrieval.