Technical Specifications of Gemini 3.7 Flash
| Item | Gemini 3.7 Flash |
|---|---|
| Model ID | gemini-3.7-flash |
| Provider | |
| Context | 1,048,576 tokens |
| Max Output | 65,536 tokens |
| Input | Text / Image / Video / Audio / PDF |
| Thinking | Low / Medium / High |
| Function Calling | ✅ |
| Code Execution | ✅ |
| Search Grounding | ✅ |
| Google Maps Grounding | ✅ |
| URL Context | ✅ |
| File Search | ✅ |
| Computer Use | Preview |
| Knowledge Cutoff | March 2026 |
| Release | Aug. 13, 2026 |
What Is Gemini 3.7 Flash?
Gemini 3.7 Flash is Google's August 13, 2026 iteration of the Gemini 3 Flash line. Google positions it as its most intelligent Flash workhorse yet for coding and AI agents, with improvements focused on software engineering, web development, complex knowledge work, and multi-step workflows. It arrived three weeks after Gemini 3.6 Flash and is described as the result of developer feedback and algorithmic innovations.
Main Features
- 1M-token context: Supports large repositories, long documents, extended conversations, and multi-step agent histories.
- Multimodal reasoning: Accepts text, images, video, audio, and PDF inputs.
- Configurable thinking: Supports LOW, MEDIUM, and HIGH thinking;
MINIMALis not supported. - Agentic coding: Major gains in debugging, issue resolution, first-pass code accuracy, and long-horizon software engineering.
- Web and UI development: Stronger adherence to screenshots, images, and design systems.
- Tool-driven workflows: Function calling, code execution, Search grounding, Google Maps grounding, URL context, file search, and preview computer use.
- Document-heavy knowledge work: Improved performance in finance, law, biosciences, PDF comprehension, and enterprise automation.
Benchmark Performance
| Benchmark | Gemini 3.7 Flash | Gemini 3.6 Flash | Claude Sonnet 5 | GPT-5.6 Terra |
|---|---|---|---|---|
| Artificial Analysis Intelligence Index | 56 | 52 | 55 | 57 |
| FrontierCode 1.1 Main | 43.6% | 34.4% | 42.7% | 41.3% |
| DeepSWE v1.1 | 65.3% | 49.0% | — | — |
| GDM-MRCR v2 (8-needle, 128K) | 97.0% | 91.8% | 81.5% | 93.5% |
| OSWorld-2.0 | 47.9% | 33.8% | — | 50.2% |
| Agent's Last Exam | 26.3% | 24.2% | 33.3% | 28.0% |
| HLE-Verified | 53.6% | 51.2% | 31.0% | 51.1% |
| LABBench2 | 82.1% | 76.1% | 80.1% | 81.2% |
Google also reports GDP.pdf at 34.0% versus 22.0%, AutomationBench at 30.4% versus 17.0%, and WebDev Arena Elo at 1588 versus 1538 for Gemini 3.6 Flash.
Gemini 3.7 Flash vs Gemini 3.6 Flash vs Claude Sonnet 5
| Model | Positioning | Context | Key strengths |
|---|---|---|---|
| Gemini 3.7 Flash | Efficient agentic workhorse | 1.05M | Coding, web development, multimodal agents, long-context workflows |
| Gemini 3.6 Flash | Previous Flash workhorse | 1M | General agent and multimodal workloads |
| Claude Sonnet 5 | Higher-end reasoning/coding | Varies | Coding, knowledge work, reasoning |
Gemini 3.7 Flash is particularly attractive when an application needs a balance between frontier capability, multimodal processing, long context, tool use, and inference economics.
Limitations
MINIMALthinking is unavailable.- Image generation is not supported; output is text.
- Gemini Live API is not supported.
- Knowledge cutoff is March 2026.
- Google notes possible hallucinations, occasional slowness, and timeout issues.
- Computer use is currently a Preview capability.
Use Cases
- Coding agents and repository-scale software engineering.
- Web development from screenshots, images, or design systems.
- Enterprise document and knowledge agents.
- Multimodal workflow automation.
- Tool-using agents with search, code execution, and function calling.
- Long-context analysis of large codebases and document collections.
- Structured enterprise automation.