What Gemini 3.7 Flash is
Where 3.7 Flash sits
Gemini 3.7 Flash is the newest Flash-series model, released by Google in August 2026.
The flagship model for coding and agents
Google positioned this model as its most intelligent workhorse yet for coding and agents. "Workhorse" does not mean competing at the very top of raw intelligence; it means the model that handles large volumes of everyday work. An agent, here, is a way of using AI in which it carries out multiple steps on its own rather than being instructed at each one.
"Today, we’re building on the progress of our widely used Flash series by introducing Gemini 3.7 Flash, our most intelligent workhorse model yet for coding and agents." — from the Google official blog (opening)
Released just three weeks after 3.6 Flash
3.7 Flash arrived three weeks after 3.6 Flash shipped. Google calls this a direct result of developer feedback and algorithmic innovations, which it says it will bring to future models. For where the previous generation sits, see the 3.6 Flash explainer.
"This release comes just three weeks after Gemini 3.6 Flash, and is a direct result of developer feedback and algorithmic innovations that we look forward to bringing to future models." — from the Google official blog (introduction)
How it differs from 3.6 Flash
The two metrics with the largest gains (%)
DeepSWE gained 1.3x and Automation 1.8x, so the larger relative gain is in Automation. But its absolute score is still 30.4 — automation tasks fail two times in three.
The five metrics Google published are as follows.
| Metric | 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 (Elo) | 1588 | 1538 |
| GDP.pdf | 34.0% | 22.0% |
| AutomationBench | 30.4% | 17.0% |
"It also achieves higher first-pass code accuracy and has improved performance in generating production-ready code as seen in FrontierCode 1.1 Main (43.6% vs 34.4%) and DeepSWE v1.1 (65.3% vs 49.0%)."(coding section)／"It outperforms 3.6 Flash on Arena.ai’s WebDev Arena with an Elo score of 1588 vs 1538."(web development section)／"It significantly outperforms 3.6 Flash on the GDP.pdf benchmark (34.0% vs 22.0%)"／"It also surpasses 3.6 Flash in AutomationBench, demonstrating it can more effectively complete real-world business workflows (30.4% vs 17.0%)."(specialized fields section) — from the Google official blog
Coding performance (DeepSWE and FrontierCode)
DeepSWE v1.1 is 65.3% for 3.7 Flash against 49.0% for 3.6 Flash, and FrontierCode 1.1 Main is 43.6% against 34.4%. What Google points to is better debugging and issue resolution, plus first-pass code accuracy.
"3.7 Flash shows strong gains over 3.6 Flash in coding tasks like debugging and issue resolution." — from the Google official blog (coding section)
Web development, document processing, and business automation
WebDev Arena Elo goes from 1538 to 1588. GDP.pdf, which covers specialized documents, rises from 22.0% to 34.0%, and AutomationBench, which measures whether a model can complete business workflows, rises from 17.0% to 30.4%. Elo is a rating that expresses relative strength. The largest proportional gain is AutomationBench at roughly 1.8×, while in raw point difference DeepSWE is the largest.
"For knowledge-dense fields like finance, law, and biosciences, 3.7 Flash delivers improved reasoning and accuracy."(specialized fields section)／"an eval for testing a model’s ability to process complex documents"(description of GDP.pdf) — from the Google official blog
Pricing and where to use it
Pricing (per 1M tokens)
The price has an expiry date.
Half price through the end of the year
Through the end of 2026, 3.7 Flash is available at an introductory price of $0.75 per 1M input tokens and $3.75 per 1M output tokens — half the cost of 3.6 Flash. A token is the unit a text is split into for processing, and pricing is based on that volume. The introductory pricing ends on December 31, 2026, and from January 1, 2027 it returns to $1.50 input and $7.50 output. Since the unit price doubles at the turn of the year, budgeting at the regular price is the safer basis if you plan to run it heavily in production.
"3.7 Flash is available through the end of the year at an introductory price of $0.75/1M input tokens and $3.75/1M output tokens."(pricing section)／"3.7 Flash delivers substantial improvements across software engineering, knowledge work, and web development workflows — with an introductory price of half the original 3.6 Flash cost per million tokens."(introduction)／"Introductory pricing expires on December 31, 2026. Starting January 1, 2027, $1.50/1M input tokens and $7.50/1M output tokens will apply."(footnote) — from the Google official blog
Where it is available, and the takeaway
Developers can use it in Google Antigravity or the Gemini API (AI Studio and Android Studio); enterprises can use it in the Gemini Enterprise Agent Platform and the Gemini Enterprise app. On the consumer side, Gemini Spark — available to AI Pro and Ultra subscribers in over 160 countries — switched to 3.7 Flash on August 13, the same day as the announcement. The gains are largest on multi-step work, which makes it worth revisiting processes you had been stitching together by hand.
"Developers: Explore agent-first workflows in Google Antigravity or start building today in the Gemini API via Google AI Studio and Android Studio."／"Enterprises: Access 3.7 Flash in Gemini Enterprise Agent Platform and the Gemini Enterprise app."(both from the Try it today section)／"Gemini Spark, available to Google AI Pro and Ultra subscribers in over 160 countries, will be using Gemini 3.7 Flash starting today."(Spark section) — from the Google official blog
For the wider picture, see the Gemini 3.5 guide. To convert the announcement page into markdown for reading, the following tool can help.
