Gemini 3 Flash Expands Across Google Services For Faster Multimodal AI

Google is expanding the Gemini 3 family with Gemini 3 Flash, a leaner version of its flagship AI model aimed at faster and cheaper use across products. The upgrade follows the original Gemini 3 launch in November 2025 and focuses on multimodal reasoning, agent-like behaviour and quick responses for everyday work, while keeping performance at a high level for complex tasks.

Gemini 3 Flash targets frontier-grade intelligence while trimming infrastructure costs, using about 30% fewer tokens on average than earlier Flash models. It handles real-time replies for demanding prompts, spanning video understanding, audio processing and planning workflows. Benchmark tests show progress in specialist areas such as software development, where the model improves results on the SWE-bench coding benchmark and delivers stronger agentic coding support for users worldwide.

Gemini 3 Flash AI model features and availability

Google is rolling out Gemini 3 Flash across its main services, with the model now running AI Mode in Google Search by default and taking over from Gemini 2.5 Flash in the Gemini app for Pro and Ultra subscribers. Developers can tap the model through Vertex AI, AI Studio and Gemini CLI, while enterprise customers, including Salesforce and Figma, are already adopting it for their own products and internal tools.

Alongside Gemini 3 Flash, Google is testing a Gemini 3 Deep Think mode with a limited group of users after safety checks, designed to push reasoning depth beyond the main release while the current Flash model broadens access to advanced AI capabilities through consumer apps, developer tools and enterprise platforms.