Gemini 3.7 Flash
Gemini 3.7 Flash is Google DeepMind’s flagship workhorse model engineered for high-throughput, low-latency reasoning, autonomous agent workflows, and multi-file code synthesis. Positioned as the performance-optimized speed tier in the Gemini 3 family, 3.7 Flash delivers frontier-class intelligence and multi-modal understanding at a fraction of the cost and latency of Pro-tier engines.
Model Overview
Gemini 3.7 Flash introduces major benchmark jumps in agentic execution, browser automation, and multi-step problem solving compared to previous Flash generations. It serves as a primary intelligence layer for enterprise automation, real-time developer workflows, and autonomous sub-agent systems.
- Developer: Google DeepMind
- Model Class: High-Efficiency Multimodal Reasoning Model
- Release Date: August 13, 2026
- Primary Target Use Cases: Autonomous agentic workflows, low-latency coding/debugging, desktop and browser automation (Computer Use), document/video reasoning, and high-volume API routing.
Core Capabilities & Performance Profile
Accelerated Agentic & Software Engineering
- Substantial Benchmark Gains: Demonstrates major performance jumps over 3.6 Flash across agentic coding (DeepSWE v1.1) and task automation (AutomationBench).
- Long Output Generation: Supports up to 64,000 output tokens in a single completion, enabling full codebase refactoring, multi-file code rewrites, and lengthy structured documentation without truncation.
Dynamic Thinking Controls
- Tunable Chain-of-Thought: Features explicit controls for reasoning depth via the
thinking API configuration (low, medium, high). This allows developers to balance speed against logical rigor depending on prompt complexity.
Multi-Modal Context & Native Computer Use
- 1M Token Input Window: Ingests massive contexts including dense PDF documents, high-frame-rate video streams, native audio, and complex visual layouts.
- Native Computer & Browser Use: Integrates direct support for driving visual interfaces, parsing UI elements, executing desktop/web browser clicks, and navigating terminal environments.
Technical Specifications
| Parameter |
Specification |
| Model ID |
gemini-3.7-flash |
| Input Context Window |
1,048,576 tokens (1M) |
| Max Output Tokens |
65,536 tokens (64k) |
| Input Modalities |
Text, Code, Images, Audio, Video, PDF |
| Output Modalities |
Text, Code, Structured JSON |
| Reasoning Configuration |
Configurable Thinking (low, medium, high) |
| Native Capabilities |
Computer Use, Search Grounding, Maps Grounding, Code Execution, Function Calling |
Availability & Access Channels
- Developer Access: Available via Google AI Studio, Vertex AI, Google Antigravity, and Gemini Cloud API endpoints.
- Agent Integration: Compatible with terminal agent tools, CLI execution runtimes, and multi-agent sub-task pipelines.
*