Google has introduced Gemini 3.7 Flash, positioning the new model as its most capable Flash model yet for coding and agent-based workloads. The release arrives only three weeks after Gemini 3.6 Flash, an unusually short turnaround that Google says reflects developer feedback and new algorithmic improvements.
Rather than framing 3.7 Flash simply as another incremental model update, Google is focusing on a combination that matters to developers deploying AI at scale: stronger performance across coding, complex documents and business workflows, alongside a lower introductory price.
The company says Gemini 3.7 Flash will be available through the end of 2026 at $0.75 per 1 million input tokens and $3.75 per 1 million output tokens. Google describes the introductory pricing as half the original cost of Gemini 3.6 Flash per million tokens.
Coding gains form a major part of the 3.7 Flash pitch
Software engineering is one of the clearest areas where Google is claiming measurable improvement.
On FrontierCode 1.1 Main, which Google cites as an evaluation of production-code quality, Gemini 3.7 Flash recorded a score of 43.6%, compared with 34.4% for Gemini 3.6 Flash.
The gap was wider on DeepSWE v1.1, where the new model scored 65.3% against 49.0% for its predecessor. Google says the model has also improved in areas such as debugging, issue resolution and first-pass code accuracy.
For web development, Google says Gemini 3.7 Flash can produce more functional layouts and feature-complete applications with fewer prompts. The model recorded an Elo score of 1,588 on Arena.ai’s WebDev Arena, up from 1,538 for Gemini 3.6 Flash.
That improvement is particularly relevant to Google’s broader effort to make Flash useful for workflows where a model is expected to do more than answer a prompt. The company is pitching 3.7 Flash as a model capable of planning, using tools and working through multi-step tasks with less manual intervention.
The less obvious jump is in business workflows
One of the more notable figures in Google’s benchmark disclosures comes from AutomationBench, which evaluates the completion of real-world business workflows.
Gemini 3.7 Flash scored 30.4%, compared with 17.0% for Gemini 3.6 Flash. Google says the result reflects an improved ability to carry out practical multi-step tasks rather than simply generate text.
The company is also highlighting stronger performance in knowledge-heavy fields including finance, law and biosciences.
On the GDP.pdf benchmark, designed to test a model’s ability to work through complex documents, Gemini 3.7 Flash scored 34.0%, up from 22.0% for 3.6 Flash.
Taken together, those numbers show where Google is trying to move the Flash family. Speed and cost remain central to the product, but the company is increasingly positioning Flash for tasks that involve reasoning across documents, software environments and connected tools.
Google says developers should need fewer retries
Beyond benchmark scores, Google says it has changed how Gemini 3.7 Flash behaves when working through longer assignments.
According to the company, the model is better at adapting when it encounters roadblocks, clarifying intent when necessary and following instructions more faithfully. Google also says 3.7 Flash puts more effort into multi-step planning and tool calls, which is intended to reduce retries and the amount of manual oversight required in engineering workflows.
That distinction matters for agent-style applications. A model used inside an automated workflow has to do more than provide a strong isolated answer. It needs to stay on task while handling several steps, deciding when to use tools and recovering when part of a workflow does not proceed as expected.
Google is making that agent-oriented use case one of the central themes of Gemini 3.7 Flash.
Gemini Spark is moving to 3.7 Flash
The new model is also being brought into Gemini Spark, Google’s personal AI agent for Google AI Pro and Ultra subscribers.
Google says Spark is available in more than 160 countries and will begin using Gemini 3.7 Flash. The company says the change is intended to improve knowledge-work tasks and tool use across Google Workspace applications, particularly for more complicated workflows requiring several different capabilities.
With the new model, Google says Spark can more efficiently handle work such as consolidating files, drafting emails and updating status documents.
That integration also gives Google a direct consumer-facing route for deploying the capabilities it is emphasizing in 3.7 Flash, rather than limiting the release to developers using the Gemini API.
Google adds updated safeguards for cyber and CBRN misuse
Google says Gemini 3.7 Flash is shipping with updated safeguards covering misuse involving chemical, biological, radiological and nuclear risks, as well as cyber offence.
The company says these protections are part of its broader Frontier Safety safeguards and are intended to restrict harmful uses while continuing to support beneficial applications.
Where Gemini 3.7 Flash is available
Developers can access Gemini 3.7 Flash through the Gemini API via Google AI Studio and Android Studio, as well as use it for agent-focused workflows in Google Antigravity.
Enterprise customers can access the model through the Gemini Enterprise Agent Platform and Gemini Enterprise app.
For individuals, the model is being made available through Spark in the Gemini app for Google AI Pro and Ultra subscribers in supported countries.
The release leaves Google making a fairly specific bet with the Flash line: that developers increasingly want a model that can handle substantial coding and agentic workloads without moving into a higher-cost tier for every task.
With Gemini 3.7 Flash arriving only three weeks after its predecessor, the pace of that effort is becoming almost as notable as the benchmark gains themselves.
Also Read: Apple Begins iPhone 18 Pro Production in India for Global Exports
