Agents

Google details Gemini 3.8 Flash with up to 1M-token context window

The model accepts text, image, audio and video inputs and is positioned as an upgrade for software-engineering and agentic knowledge workflows.