Key Takeaways
Gemini 3.7 Flash is Google's newest workhorse model, launched on August 13, 2026, built specifically for coding, agentic workflows, and enterprise automation rather than general chatbot use.
It launched just three weeks after Gemini 3.6 Flash — one of Google's fastest model turnarounds — and replaced it as the default Flash-tier model.
It's priced at an introductory rate of $0.75/1M input tokens and $3.75/1M output tokens, roughly half of standard 3.6 Flash pricing, available only through December 31, 2026.
On independent benchmarks, 3.7 Flash beats 3.6 Flash by wide margins — including a jump from 34.4% to 43.6% on FrontierCode and 49.0% to 65.3% on DeepSWE for software engineering tasks.
It's distilled from Google's broader Gemini research, optimized to run efficiently for high-volume, production-grade agent and coding workloads rather than as a standalone flagship model.
It now powers Gemini Spark, Google's 24/7 personal AI agent, improving tool use across Google Workspace apps for Pro and Ultra subscribers.
This is a live, day-one-usable model — accessible immediately via Google AI Studio, the Gemini API, Android Studio, Google Antigravity, and the Gemini Enterprise Agent Platform, not just an announcement.
Gemini 3.7 Flash is Google's newest open, API-accessible AI model, released on August 13, 2026, as an upgrade to Gemini 3.6 Flash. It's optimized for coding, agentic tool use, and enterprise document workflows, and it's available at an introductory price of $0.75/1M input tokens and $3.75/1M output tokens — about half the cost of 3.6 Flash — through the Gemini API, Google Cloud, and the Gemini app.
This guide breaks down everything you need to know about Gemini 3.7 Flash: what it is, how it performs against Gemini 3.6 Flash on real benchmarks, its pricing structure, and exactly how to access it through the API.
What Is Gemini 3.7 Flash?
Gemini 3.7 Flash is the newest entry in Google's Flash model family, built specifically to handle software engineering, multi-step agent workflows, and knowledge-intensive document work. Unlike Google's larger "Pro" tier models, the Flash series is engineered for a balance of speed, cost-efficiency, and strong reasoning — making it the go-to choice for developers building production applications, autonomous agents, and high-volume enterprise tools.
According to Google, the model was developed largely in response to direct developer feedback following the 3.6 Flash release, combined with internal algorithmic improvements the team plans to carry forward into future model generations.
Gemini 3.7 Flash Release Date
Gemini 3.7 Flash officially launched on August 13, 2026. The announcement was made through Google's official blog by Tulsee Doshi, Senior Director of Product Management for the Gemini team, on behalf of Google DeepMind's Gemini division.
This release cadence is worth noting: Gemini 3.6 Flash had only been available for about three weeks before 3.7 Flash replaced it as Google's leading Flash-tier model. This rapid iteration cycle reflects the intensifying competition in the AI model space, where labs are pushing frequent, incremental upgrades rather than waiting for large, infrequent version jumps.
The model became available immediately upon announcement across Google's developer and enterprise platforms, as well as inside Gemini's consumer-facing agent product.
Key Improvements in Gemini 3.7 Flash
Google highlights three major areas of improvement with this release:
1. Stronger Coding and Software Engineering Ability
Gemini 3.7 Flash shows meaningful gains in debugging, issue resolution, and generating code that is ready for production use on the first attempt. This matters for development teams that rely on AI-assisted coding tools, since fewer errors on the initial pass translates directly into less manual review and faster shipping.
2. Better Web and UI Development
The model has been tuned to generate more complete, functional web layouts and applications using fewer prompts. It's also better at matching a design reference — whether that's a screenshot, an uploaded image, or an existing design system — a capability that's especially valuable for teams doing rapid prototyping or converting mockups into working front-end code.
3. Improved Reasoning for Knowledge-Dense Work
For fields like law, finance, and the biosciences — where documents are dense, technical, and context-heavy — Gemini 3.7 Flash demonstrates notably better comprehension and reasoning accuracy compared to its predecessor. This makes it more viable for enterprise use cases involving contract review, financial analysis, and complex document processing.
Google also emphasizes that the model exhibits more disciplined, deliberate behavior during multi-step agent tasks. It plans further ahead, uses tools more effectively, adapts better when it encounters obstacles, and asks for clarification when a task is ambiguous — reducing the need for constant human oversight during long agentic workflows.
Gemini 3.7 Flash Benchmark Performance
Benchmarks are where Gemini 3.7 Flash makes its strongest case. Based on figures released by Google, here's how the model performed on independent evaluation suites compared to Gemini 3.6 Flash:

Image Credit Gemini
Benchmark | What It Measures | Gemini 3.6 Flash | Gemini 3.7 Flash |
FrontierCode 1.1 Main | Production-quality code generation | 34.4% | 43.6% |
DeepSWE v1.1 | Long-horizon software engineering tasks | 49.0% | 65.3% |
WebDev Arena (Elo) | Real-world web development quality | 1538 | 1588 |
GDP.pdf | Complex document/PDF comprehension | 22.0% | 34.0% |
AutomationBench | Real-world business workflow automation | 17.0% | 30.4% |
A few of these evaluations come from independent or third-party organizations rather than Google itself, which adds credibility to the reported gains:

FrontierCode 1.1 is maintained by Cognition, the company behind the Devin AI software engineer.

DeepSWE v1.1 is run by DataCurve.

WebDev Arena rankings are published on Arena.ai's public leaderboard, which crowdsources human preference votes on AI-generated web code.

AutomationBench was introduced by Zapier to specifically test how well AI models can complete real business automation tasks.
The consistent theme across these benchmarks: Gemini 3.7 Flash isn't a marginal update — in several categories, it nearly doubles the performance of the model it replaces after only a few weeks in the market.
Gemini 3.7 Flash vs Gemini 3.6 Flash: What Actually Changed?
Since these two models launched so close together, it's natural to ask whether upgrading is worth it. Here's a side-by-side breakdown:
Category | Gemini 3.6 Flash | Gemini 3.7 Flash |
Release timing | ~3 weeks before 3.7 Flash | August 13, 2026 |
Coding accuracy (FrontierCode) | 34.4% | 43.6% (↑ ~27%) |
Software engineering (DeepSWE) | 49.0% | 65.3% (↑ ~33%) |
Web development (WebDev Arena Elo) | 1538 | 1588 |
Document comprehension (GDP.pdf) | 22.0% | 34.0% |
Business workflow automation | 17.0% | 30.4% |
Introductory price (input) | Standard 3.6 Flash rate | 50% lower than 3.6 Flash |
Agentic tool use | Baseline | More disciplined, fewer retries |
Powers Gemini Spark (consumer agent) | Previous version | Yes, updated as of launch |
In short: Gemini 3.7 Flash outperforms 3.6 Flash across nearly every measured category, while launching at half the introductory per-token cost. For developers already on 3.6 Flash, this makes 3.7 Flash a fairly clear upgrade path, especially for coding and agent-heavy workloads.
Gemini 3.7 Flash Pricing
One of the more surprising elements of this release is the pricing strategy. Google is offering Gemini 3.7 Flash at an introductory rate of $0.75 per 1 million input tokens and $3.75 per 1 million output tokens — roughly half the cost of the standard 3.6 Flash pricing.
However, this is a limited-time offer. According to the official announcement, this introductory pricing is valid only through December 31, 2026. Starting January 1, 2027, pricing will increase to $1.50 per 1 million input tokens and $7.50 per 1 million output tokens.
For teams building at scale, this pricing window is worth factoring into budget planning — locking in usage patterns before the price increase could meaningfully affect long-term operating costs for high-volume applications.
Gemini 3.7 Flash API: How to Access It
Google has made Gemini 3.7 Flash available across its full developer and enterprise ecosystem:
For developers: The model can be accessed through the Gemini API via Google AI Studio, and it's also integrated into Android Studio. Google has published a full developer guide covering setup and implementation. Developers focused on agent-first workflows can also explore Google Antigravity, Google's platform built specifically for orchestrating AI agents.
For enterprises: Gemini 3.7 Flash is available through the Gemini Enterprise Agent Platform inside Google Cloud's Model Garden, as well as through the standalone Gemini Enterprise application.
For individual users: The model powers Gemini Spark, Google's always-on personal AI agent, which is bundled with Google AI Pro and Ultra subscriptions in over 160 supported countries.
Getting started via the API is straightforward for anyone already familiar with Gemini's developer tools — the model slots into the existing Gemini API structure, so migrating from 3.6 Flash to 3.7 Flash typically requires only a model-name change in existing integrations, rather than a full rebuild.
Real-World Adoption
Google notes that early feedback from companies testing Gemini 3.7 Flash has been positive, citing measurable improvements in both precision and cost efficiency compared to the previous model. Organizations referenced in the announcement as early adopters or testers include Box, Browser Use, Cartwheel, Databricks, Harvey, Hebbia, LangChain, Pydantic, and Stanford's Department of Biology, among others — spanning industries from legal tech and enterprise data to academic research and developer tooling.
This breadth of adoption suggests Gemini 3.7 Flash isn't positioned as a narrow coding tool, but rather as a general-purpose workhorse model suited to a wide range of professional and technical workflows.
Powering Gemini Spark
Beyond the developer-facing API, Gemini 3.7 Flash now also drives Gemini Spark, Google's 24/7 personal AI agent that was first introduced at Google I/O. With the 3.7 Flash update, Spark gains improved tool-use capabilities across Google Workspace apps, allowing it to more reliably consolidate files, draft emails, and manage status updates on a user's behalf — with better accuracy on complex, multi-skill tasks that require chaining several actions together.
Safety and Responsible Deployment
As with previous Gemini releases, Google states that 3.7 Flash ships with updated safety measures aligned with its Frontier Safety Framework. The company specifically mentions strengthened safeguards in two sensitive domains: Chemical, Biological, Radiological, and Nuclear (CBRN) misuse and cyber offense risks, while still aiming to preserve legitimate, beneficial use cases in these fields. Full technical details are available in the model's official safety and bioresilience documentation and its published model card.
Final Thoughts
Gemini 3.7 Flash represents one of Google's fastest model turnarounds to date, and the benchmark data backs up the claim that it's a genuine leap rather than an incremental patch. With lower introductory pricing, stronger coding and agentic performance, and broader availability across developer, enterprise, and consumer products, it's positioned to become the default choice for teams building AI-powered coding tools and automation agents through the rest of 2026.
For teams currently on Gemini 3.6 Flash, the combination of better benchmark performance and a temporarily lower price point makes this a strong window to evaluate an upgrade — particularly before the pricing shift scheduled for January 2027.
FAQs
When was Gemini 3.7 Flash released?
Gemini 3.7 Flash was released on August 13, 2026, roughly three weeks after Gemini 3.6 Flash.
How much does Gemini 3.7 Flash cost?
The introductory price is $0.75 per 1 million input tokens and $3.75 per 1 million output tokens, available through December 31, 2026. Pricing rises to $1.50/$7.50 per 1 million tokens starting January 1, 2027.
Is Gemini 3.7 Flash better than Gemini 3.6 Flash?
Yes. It shows significant gains across coding, web development, document comprehension, and workflow automation benchmarks, in some cases nearly doubling 3.6 Flash's scores, while launching at roughly half the introductory price.
Where can I use Gemini 3.7 Flash?
Developers can access it through Google AI Studio and the Gemini API, enterprises through the Gemini Enterprise Agent Platform on Google Cloud, and individual consumers through Gemini Spark for Google AI Pro and Ultra subscribers.
What is Gemini 3.7 Flash best used for?
Based on its benchmark strengths, it's particularly well suited for software engineering and debugging, front-end/web development, agentic multi-step workflows, and processing dense technical or legal documents.
Is Gemini 3.7 Flash a replacement for Gemini 3.6 Flash?
Effectively, yes. Google positions 3.7 Flash as the new default Flash-tier model, and it now also powers the Gemini Spark consumer agent going forward.