Three weeks. That's all it took for Google to follow up Gemini 3.6 Flash with something significantly better and cheaper.
Meet Gemini 3.7 Flash. Google's calling it their "most intelligent workhorse model yet" for coding and agent-based workflows. And they're backing that claim with numbers.
What's Actually Improved?
Let's cut through the jargon. Here's what 3.7 Flash does better:
Coding and software engineering. The model scores 43.6% on FrontierCode 1.1 Main—a benchmark for production-ready code compared to 3.6 Flash's 34.4%. On DeepSWE v1.1, which tests long-horizon software engineering, it jumps from 49.0% to 65.3%. Fewer bugs, cleaner code, less debugging.
Web development. WebDev Arena scores tell the story: 1588 Elo for 3.7 Flash versus 1538 for 3.6 Flash. The new model generates more functional layouts and feature-complete apps in fewer prompts. Feed it a screenshot or a design system, and it matches the reference input with high accuracy.
Knowledge work. For finance, law, and biosciences, the improvements are just as striking. On the GDP.pdf benchmark which tests complex document comprehension 3.7 Flash scores 34.0% versus 22.0%. In AutomationBench, it handles real-world business workflows at 30.4% compared to 17.0%.
Better Developer Experience and Price
Beyond benchmarks, developers will notice a smoother experience. The model adapts to roadblocks, clarifies intent when it's unclear, and follows instructions with greater precision. It thinks more carefully before acting fewer retries, less manual oversight.
And here's the kicker: it costs half the price of 3.6 Flash. Introductory pricing is set at $0.75 per million input tokens and $3.75 per million output tokens available through the end of the year.
What Early Customers Are Saying
The customer list reads like a who's who of AI adoption: Box, Databricks, Harvey, LangChain, Pydantic, and Stanford's Department of Biology, among others. Early feedback highlights the model's performance and precision at a surprisingly low cost.
Gemini Spark Gets an Upgrade Too
Gemini Spark Google's personal AI agent available to Pro and Ultra subscribers in over 160 countries is now powered by 3.7 Flash. The upgrade improves tool use for Google Workspace apps, making it more efficient for knowledge work. Think consolidating files, drafting emails, and updating status documents with less hand-holding.
Safety Built In
Google isn't cutting corners on safety. The new model ships with updated safeguards against misuse in chemical, biological, radiological, and nuclear domains, as well as cyber offense. The full model card is available for those who want the technical details.
Where to Access It
Developers: Start building in Google AI Studio, Android Studio, or Google Antigravity.
Enterprises: Available in the Gemini Enterprise Agent Platform and Gemini Enterprise app.
Individuals: Gemini Spark users on Pro and Ultra plans get it automatically.
It's only been three weeks since 3.6 Flash. If this pace continues, the next few months are going to be very interesting.
FAQ Section
What is Gemini 3.7 Flash?
It's Google's latest AI model an upgraded version of 3.6 Flash focused on coding, software engineering, web development, and knowledge work.
How does it compare to 3.6 Flash?
Significantly better across key benchmarks: 43.6% vs 34.4% on FrontierCode (coding), 65.3% vs 49.0% on DeepSWE (software engineering), and 34.0% vs 22.0% on PDF comprehension.
Is it really half the price?
Yes. It costs $0.75 per million input tokens and $3.75 per million output tokens half the price of 3.6 Flash. This introductory pricing is available through the end of the year.
Where can I access Gemini 3.7 Flash?
Developers can use it via Google AI Studio, Android Studio, and Antigravity. Enterprises can access it through the Gemini Enterprise platform. Individual Pro and Ultra subscribers get it through Gemini Spark.
Who is already using it?
Early customers include Box, Databricks, Harvey, LangChain, Pydantic, and Stanford's Department of Biology.
