Google unveils Gemini 3.5 Flash at I/O 2026, delays Gemini 3.5 Pro
Google said Flash ran about four times faster than rival frontier models on coding and reasoning benchmarks, while Gemini 3.5 Pro was still in internal use and promised for the following month.
- Models & capabilities
- Notable
Google released Gemini 3.5 Flash at its I/O 2026 developer conference, pitching it as combining “frontier intelligence with action” for agentic and coding work rather than as a general chat model. The company said Flash ran roughly four times faster than competing frontier models while still placing at the top of coding and reasoning benchmarks, and highlighted multi-step workflows — the kind of task that might otherwise take a developer days or an auditor weeks — as the model’s intended use, shipping it into the Gemini app, Search’s AI Mode, the Antigravity coding environment and enterprise products immediately.
The larger Gemini 3.5 Pro was conspicuously absent from the launch. Google said the model was “already being used internally” but held its public release back to the following month, an unusual sequencing in which the smaller, faster model reached developers first while the flagship model remained unreleased. The company gave no technical reason for the gap, leaving open whether it reflected remaining safety evaluation, capacity constraints, or a deliberate staggering of the release.
The launch came alongside Gemini Omni, Google’s new any-input video generation model, making 3.5 Flash one of two new model families the company introduced at I/O the same day — one aimed at agentic and coding tasks, the other at generative video.