Google: Gemini 3.7 Flash

Gemini 3.7 Flash is Google's most capable Flash-tier model, released on August 13, 2026, roughly three weeks after Gemini 3.6 Flash. Google positions it as the default workhorse for agentic workflows and multimodal reasoning: it pairs frontier-level reasoning with low latency, and it ships with Computer Use in preview. The Reddit and developer communities picked it up immediately because it is also the model powering Google Antigravity's next generation of coding agents.

Key Features

• 1M-token context: 1,048,576 input tokens and 65,536 output tokens, enough for whole repositories or long research pipelines.
• Agent-first capabilities: Computer Use preview, code execution, file search, function calling, and search grounding are supported through the Gemini API.
• Multimodal input: Accepts text, images, video, audio, and PDF input, with text output.
• Reasoning controls: Thinking-token behavior is configurable per request, letting you trade latency for depth.
• Antigravity integration: Works directly with Google Antigravity for hands-on coding agent workflows.

Use Cases

• Autonomous coding agents: The Computer Use preview and code execution make it practical for agents that navigate apps, run commands, and verify their own work.
• Multimodal pipelines: PDF, image, video, and audio inputs support document understanding, visual inspection, and media-heavy research.
• Cost-conscious production: Flash-tier pricing keeps high-volume agent workloads affordable compared with Pro-tier models.
• Long-horizon tasks: The 1M context window reduces the need for aggressive summarization in complex multi-step jobs.

Pricing

Gemini 3.7 Flash has a free tier in Google AI Studio. Paid pricing is $0.75 per million input tokens and $3.75 per million output tokens (including thinking tokens) through December 31, 2026. From January 1, 2027, input rises to $1.50 per million and output to $7.50 per million. Context caching discounts are available.

Getting Started

1. Open Google AI Studio and select gemini-3.7-flash.
2. Generate an API key from Google Cloud Console.
3. Call the Gemini API with the model ID gemini-3.7-flash.
4. Enable Computer Use preview and code execution for agentic testing.

Frequently Asked Questions

Is Gemini 3.7 Flash open source?

No. It is a proprietary Google model available through Gemini products and the Gemini API.

How does Computer Use preview work?

It lets the model observe and act on a browser or desktop-like environment through the API. It is marked as preview, so production teams should test reliability and permission boundaries carefully.

What is the pricing risk?

The launch pricing is promotional: input and output costs double on January 1, 2027. Budgeting should account for the future price rather than only the launch rate.

Alternatives

• Gemini 3.5 Flash: The previous Flash-generation model, still useful for lower-cost chat workloads.
• Gemini 3 Pro: Google's Pro-tier frontier model for higher-depth reasoning.
• Claude Opus 5: Anthropic's flagship with strong agentic coding and long context.
• GPT-5.6: OpenAI's flagship reasoning family.

Tips & Best Practices

1. Budget for 2027 pricing now: Treat the December 2026 price as a temporary promotion.
2. Use thinking controls: Lower thinking effort for high-volume extraction, higher for complex planning.
3. Test Computer Use in isolation: Give the model scoped tools and inspect its actions before granting broader permissions.

Conclusion

Gemini 3.7 Flash is Google's most complete Flash-tier package for agent builders: multimodal input, 1M context, code execution, and Computer Use preview at a competitive launch price. It is a strong default candidate for new agentic products in late 2026, with the caveat that pricing rises in 2027.

Usage Terms

Usage is subject to Google's Gemini API Terms of Use. The free tier may use data to improve Google products.