Gemini Docs

Models

• New model page for Gemini 3.7 Flash (gemini-3.7-flash): multimodal Flash model for coding, agentic workflows, and multi-step execution, with 1,048,576 input / 65,536 output token limits and support for thinking, tool use, caching, Batch API, flex inference, and priority inference.
• Changelog: Gemini 3.7 Flash marked generally available (GA), with introductory pricing through December 31, 2026.
• Gemini API docs hub: Gemini 3.7 Flash added as the latest Flash model; Gemini 3.6 Flash and 3.5 Flash descriptions updated; Gemini 3.1 Flash-Lite listed.
• Deprecations: gemini-3.7-flash added (August 2026; no shutdown date announced).

Agents

• Antigravity agent and Agents: default model is now Gemini 3.7 Flash (gemini-3.7-flash); configurable models updated accordingly.
• Custom agents: examples and agent_config defaults updated to gemini-3.7-flash; supported models list includes 3.7 Flash as default.

Capabilities & inference

• Caching and Generate content — caching: Gemini 3.7 Flash and 3.6 Flash listed with a 4,096 min token limit.
• File search and Generate content — file search: Gemini 3.7 Flash marked supported.
• Flex inference: Gemini 3.7 Flash marked supported.

Examples

• CrewAI example and related agent/custom-agent samples updated to use gemini-3.7-flash.