Gemini 3.7 Flash Is Now Available on Buda
A stronger agentic workhorse for coding, web development, and multi-step tool execution.
Gemini 3.7 Flash is now available in Buda.
Google positions it as the most intelligent workhorse in the Flash family for coding and agents. In Buda, that makes it a practical route for workflows that need to write and debug code, build web interfaces, process dense documents, and complete multiple tool calls without sending every step to a heavier model.
What's new
Buda now includes Gemini 3.7 Flash in the model selector as an active chat model. It replaces Gemini 3.6 Flash in Buda's current model list.
Google released Gemini 3.7 Flash on August 13, 2026, only three weeks after Gemini 3.6 Flash. The short release interval reflects a focused update: better software engineering, web development, knowledge work, and multi-step Agent execution.
According to the official model specifications, the model accepts text, images, audio, and video, and supports a context window of up to 1,048,576 tokens with up to 65,536 output tokens. It also supports structured output, context caching, URL context, Google Search and Maps grounding, code execution, and function calling. Computer Use is available as a preview capability.
For Buda users, the important change is not the longer feature list. It is that one fast model can now carry more of an execution loop before a person or a stronger model needs to intervene.
What improved over Gemini 3.6 Flash
Google reports clear gains in coding, web development, document work, and business automation.
| Google-reported benchmark | Gemini 3.7 Flash | Gemini 3.6 Flash |
|---|---|---|
| FrontierCode 1.1 Main | 43.6% | 34.4% |
| DeepSWE v1.1 | 65.3% | 49.0% |
| WebDev Arena | 1588 Elo | 1538 Elo |
| GDP.pdf | 34.0% | 22.0% |
| AutomationBench | 30.4% | 17.0% |
These are Google's evaluations, not a guarantee for every production workflow. They point to a useful direction: better first-pass code, stronger debugging and issue resolution, closer adherence to visual references, and more reliable execution across a sequence of tools.
Google also says the model adapts better when it hits a roadblock, asks for clarification when needed, and applies more effort to multi-step planning. Those behaviors matter in Agent work because the cost of a weak model often appears as retries, manual repair, and broken handoffs rather than one visibly wrong answer.
Where Gemini 3.7 Flash fits
Use Gemini 3.7 Flash when the workflow is execution-heavy but still needs real reasoning.
- Diagnose an issue, inspect the relevant code, propose a patch, and run checks.
- Turn a screenshot or design system into a functional web interface, then revise it from feedback.
- Read long reports or mixed files and convert them into structured analysis.
- Coordinate several tool calls across Browser, Terminal, Drive, and external services.
- Run repeated research, classification, drafting, and monitoring workflows at higher volume.
- Prepare a complete first result for a person to review instead of returning fragments after every step.
It is still not the right model for every decision. High-risk legal, financial, security, or production changes need explicit review criteria and human approval. Difficult exceptions may also benefit from escalation to a deeper reasoning model.
Credits and availability
Gemini 3.7 Flash is available on Buda Free with a 0.2x Credit multiplier.
That makes it a practical option for frequent, execution-heavy Agent work. Use the right model for each step, then reserve deeper reasoning and human review for the decisions that need them.
Use Gemini 3.7 Flash in Buda
- Open the Buda dashboard and enter an Agent Workspace.
- Open the model selector in the chat input.
- Choose Gemini 3.7 Flash.
- Give the Agent the relevant files, constraints, tools, and acceptance criteria.
- Let the Agent execute the workflow, then review the result before approving consequential actions.
Buda keeps the surrounding work visible in one Agent Workspace: chat history, Drive files, Browser activity, Terminal output, and tool calls. That makes it possible to review not only what the model returned, but also what it did to get there. Learn more in the Agent Workspace guide.
A stronger execution layer still needs a manager
The most useful model update is not the one that makes people watch every step more closely. It is the one that completes more routine work cleanly while making the important review points easier to see.
Gemini 3.7 Flash gives Buda users a stronger execution route for coding, web development, document work, and multi-tool Agent workflows. The Agent can investigate, build, call tools, and assemble evidence. People still decide the goal, set the boundary, review the exceptions, and approve what matters.
Open Buda and choose Gemini 3.7 Flash for your next execution-heavy Agent workflow.