Gemini 3.7 Flash Review: Pricing, Benchmarks & API Guide 2026

Gemini 3.7 Flash is Google's Flash-tier model for coding, agent workflows and document-heavy business work. It is available from 13 August 2026 at an initial price of $0.75/1M input tokens and $3.75/1M output tokens. Now, there is a successor to the current model: Gemini 3.8 Flash, released on 2 September 2026 and it has the same price. 

In this article, we will discuss what 3.7 Flash is, its results over 3.6 Flash, its cost, its usage, and whether or not you should upgrade to 3.8 Flash.

Update: What about Gemini 3.8 Flash?

Google released Gemini 3.8 Flash, three weeks after 3.7, which makes it the third Flash model in about six weeks. It's based on the same 3.7 Flash—not a new base, and the list price is the same. What has changed is their usage: it does more reasoning and tool use per thing, so the same rate card still adds up to a larger bill.

Gemini 3.7 Flash at a glance

• Released: 13 August 2026, when it replaced 3.6 Flash as the lead Flash-tier model at Google.

• Built for: software engineering, agentic tool usage, web & UI development, dense documents.

• Price: $0.75 input / $3.75 output per 1M tokens until 31 December 2026. Subsequently: $1.50 / $7.50.

• Context: ~ 1M input tokens and approx. 64K output tokens.

• Availability: Google AI Studio, Gemini API, Android Studio, Google Antigravity, Gemini Enterprise Agent Platform.

What is Gemini 3.7 Flash?

Google's Flash models work at the top-end depth for speed and cost. They're the ones teams put behind apps, agents and high-volume tools, while the Pro models handle the hardest reasoning. Gemini 3.7 Flash follows that pattern. Google says it was shaped largely by developer feedback on 3.6 Flash, plus the internal algorithm changes the plans to carry into later models.

What improved over Gemini 3.6 Flash?

Coding. Google points to better debugging, issue resolution and code that's closer to production-ready on the first try. That matters more than it sounds. 

Web and UI work. The model builds fuller layouts with fewer prompts and follows a design reference better, whether that's a screenshot, an image, or an existing design system. Useful for prototyping and for turning mockups into front-end code.

Dense documents. Law, finance and biosciences are the examples Google gives. Contracts, financial filings and technical papers are the kind of material where the reasoning gap between 3.6 and 3.7 shows up.

Agent behavior. Google also says the model plans further ahead, uses tools more sensibly, recovers better from obstacles and asks for clarification when a task is ambiguous. That should mean less babysitting on long multi-step runs.

Gemini 3.7 Flash benchmarks

These are the figures Google published against 3.6 Flash

The jumps are real but uneven. AutomationBench led the gains with a near 80% surge from a low baseline, while DeepSWE and FrontierCode rose by approximately one-third and one-quarter, respectively. Additionally, Box released its enterprise knowledge-work evaluation, showing that 3.7 Flash outperformed 3.6 Flash with a score of 67% compared to 62%. 

Benchmarks are a guide, not a verdict. Run the model on your own prompts before you decide.

Gemini 3.7 Flash vs 3.6 Flash vs 3.8 Flash

The practical read: moving from 3.6 to 3.7 is an easy yes for coding and agent work. Moving from 3.7 to 3.8 depends on whether the extra reasoning pays for itself on your tasks. Test both on the same prompts and compare the outputs, not just the rate card.

How much does Gemini 3.7 Flash cost?

First, the output price includes thinking tokens, so a model that reasons a lot costs more per task than the rate suggests. Second, the price doubles on 1 January 2027, so any budget for next year should use the higher numbers.

Google also lists a free tier on the Gemini API for Flash models, with rate limits. Limits and eligibility change, so confirm them on Google's pricing page.

Technical specs worth knowing

The model accepts text, images, audio, video and PDFs, with a context window of roughly one million input tokens and an output cap of about 64K tokens. Thinking is configurable at Low, Medium (the default) and High. The Minimal setting isn't supported. If you're tuning cost, the thinking level is the first dial to turn.

How to access Gemini 3.7 Flash through the API

Developers: use the Gemini API through Google AI Studio, or work in Android Studio. For agent-first projects, Google Antigravity is built to orchestrate agents.

Enterprises: the model is available through the Gemini Enterprise Agent Platform in Google Cloud's Model Garden, and through the Gemini Enterprise app.

Individuals: the model powers Gemini Spark, Google's always-on personal agent, at launch. Spark comes with Google AI Pro and Ultra subscriptions.

If you're already on 3.6 Flash, migrating is usually just a model-name change in your existing integration. The model ID for this release is gemini-3.7-flash, and the newer one is gemini-3.8-flash. Test before you flip production traffic, since behavior can differ even when the code doesn't.

Who should use Gemini 3.7 Flash?

It fits teams building coding assistants, UI generators, document-review tools and multi-step agents where speed and cost matter. It's also a sensible default for high-volume jobs where a Pro-tier model would be overkill.

It's a weaker fit when a task needs deep, open-ended planning with no human check. Some community users report that Flash-tier models sometimes take the direct, simple route on complex architecture work. That's anecdotal, so test it on your hardest task before you rely on it.

Who has tried it?

Google names Box, Browser Use, Cartwheel, Databricks, Harvey, Hebbia, LangChain, Pydantic and Stanford's Department of Biology among early testers. The mix runs from legal tech to enterprise data to academic research, which suggests Google wants this to be a general workhorse and not a narrow coding tool.

Safety

Google says the release follows its Frontier Safety Framework, with extra safeguards around chemical, biological, radiological and nuclear misuse and around cyber offense. The model card and safety documentation carry the full details.

Final verdict

Gemini 3.7 Flash was a real step up from 3.6 Flash, especially for coding and agents, and it launched at a price that's hard to argue with. Its position has moved, though. Gemini 3.8 Flash now sits above it at the same price, so the useful question isn't whether 3.7 is good. It's whether 3.8 does enough extra for your workload to justify the extra tokens. Try on your own work and budget for the January 2027 price increase.

Need assistance in selecting a model, creating an AI agent, or integrating Gemini into your product? It's built by our AI Development team, and our AI Consulting team can help you select the right setup initially. Schedule a Free Consultation.

To learn more about the future of generative AI and the trends to watch, check out The Future of Generative AI: 8 Key Trends.

Frequently asked questions

When was Gemini 3.7 Flash released? 

Approximately 3 weeks after the launch of Gemini 3.6 Flash, on 13 August 2026.

What is the price of Gemini 3.7 Flash?

$0.75 per 1M input tokens and $3.75 per 1M output tokens until 31 December 2026. From 1 January 2027, the price rises to $1.50 and $7.50.

Is Gemini 3.7 Flash free?

Google has a free tier on the Gemini API for the Flash models, with rate limits. The above charges for paid use apply per token. Please consult Google's pricing page to find the current limits.

Is Gemini 3.7 Flash better than 3.6 Flash?

Yes, on published benchmarks. It performs better in coding, web development, document comprehension, or workflow automation, and Google claims it performs better when tasks are multi-step ones with agents.

What are the differences in the versions of Gemini 3.7 Flash and 3.8 Flash?

Here are the changes between 3.8 Flash and 3.7 Flash: Its list price is the same. In 3.8, the number of reasoning and tool-use task items has increased, increasing the cost of a completed task.

Can I use other versions of Flash instead of 3.7?

It is called the "old" Flash model by Google and some sources refer to it as "old" or "deprecated," while others indicate no shutdown date. Please read the model list and changelog on Google.

What is Gemini 3.7 Flash best for?

Software engineering and debugging, front-end and web development, multi-step agents, and dense legal, financial, or technical documents.

How to use the Gemini 3.7 Flash API?

Use the model ID gemini-3.7-flash in Google AI Studio using the Gemini API or in Vertex AI or with the Gemini Enterprise Agent Platform.