Gemini Flash

Google has launched Gemini 3.7 Flash, a new model designed specifically for coding, AI agents and complex multi-step workflows. Announced on August 13, 2026, the model arrives just three weeks after Gemini 3.6 Flash and brings improvements in software engineering, web development, knowledge-intensive tasks and tool use.

Google describes Gemini 3.7 Flash as its most capable “workhorse” model yet for coding and agents. Alongside higher benchmark performance, Google is emphasizing improved instruction following, planning and reliability, while offering an introductory price of $0.75 per million input tokens and $3.75 per million output tokens through the end of 2026.

What Is Gemini 3.7 Flash?

Gemini 3.7 Flash is the latest addition to Google’s Flash family, which is designed to combine relatively low latency and cost with the reasoning capabilities needed for production AI applications.

The new model targets workloads where an AI system needs to do more than generate a single response. Google is focusing on tasks involving software development, tool calls, multi-step planning, web application creation and enterprise workflows.

The release comes only three weeks after Gemini 3.6 Flash, making the version jump notable. Google says the improvements are the result of developer feedback and algorithmic changes that it expects to carry into future models.

Major Improvements in Coding

Software engineering is one of the main areas where Google says Gemini 3.7 Flash has improved.

The model is designed to perform better at debugging, issue resolution, first-pass code generation and producing production-ready software. Google reports a score of 43.6% on FrontierCode 1.1 Main, compared with 34.4% for Gemini 3.6 Flash.

On DeepSWE v1.1, Gemini 3.7 Flash scores 65.3%, compared with 49% for its predecessor. These benchmarks are intended to measure software-engineering capabilities, although benchmark results should not be treated as equivalent to performance across every real-world development environment.

The practical focus is on reducing the number of iterations developers need to get from an instruction to working code.

Better Agentic Workflows

Google is also targeting developers building AI agents.

Gemini 3.7 Flash is designed to handle longer sequences of planning and tool calls, while adapting when a workflow encounters an obstacle. Google says the model is better at recognizing when it needs clarification, following instructions accurately and continuing through multi-step tasks.

That matters for agentic systems because a model that produces a good answer in one turn can still perform poorly when it has to operate a computer, call multiple APIs, inspect results and adjust its plan.

Google says the improved execution discipline can reduce manual oversight and the need to retry failed operations.

Web Development Gets a Boost

Gemini 3.7 Flash also targets web development and UI generation.

Google says the model can create more functional layouts and feature-complete applications with fewer prompts. It can also work from reference inputs such as screenshots, images and complete design systems.

On WebDev Arena, Gemini 3.7 Flash achieved an Elo score of 1588, compared with 1538 for Gemini 3.6 Flash.

Google demonstrated the model generating interactive landing pages and coordinating sub-agents, including using Gemini Omni to create interactive visual components.

The broader implication is that Google is treating UI generation as an agentic development problem rather than simply a code-completion task.

Stronger Performance on Knowledge-Heavy Tasks

The improvements are not limited to coding.

Google reports higher performance in knowledge-intensive areas including finance, law and biosciences. On the GDP.pdf benchmark, which evaluates the ability to process complex documents, Gemini 3.7 Flash scored 34%, compared with 22% for Gemini 3.6 Flash.

It also achieved 30.4% on AutomationBench, compared with 17% for Gemini 3.6 Flash. Google uses the benchmark to measure how effectively models complete real-world business workflows.

These capabilities could be relevant to enterprise applications where an AI agent must combine document analysis, reasoning and tool use rather than simply summarize information.

Gemini 3.7 Flash Can Generate Interactive Experiences

Google is also showcasing the model as a foundation for more complex applications.

One demonstration combines Gemini 3.7 Flash with Nano Banana to dynamically generate characters, items and textures for a playable 3D game.

Another example uses the model to transform a static PDF into an interactive data story containing live charts and aggregated information. Google also demonstrated a robotics training workflow in which Gemini 3.7 Flash participates in a three-agent graph using multimodal understanding.

These examples illustrate Google’s intended direction: Flash models are increasingly being positioned as orchestration engines capable of coordinating multiple capabilities and tools.

Lower Initial Pricing

Pricing is one of the more significant parts of the announcement.

Through December 31, 2026, Gemini 3.7 Flash is available at an introductory rate of:

  • $0.75 per 1 million input tokens

  • $3.75 per 1 million output tokens

Google says the introductory price is half the original Gemini 3.6 Flash cost per million tokens.

The pricing changes on January 1, 2027. Google says the standard rate will then become:

  • $1.50 per 1 million input tokens

  • $7.50 per 1 million output tokens

The pricing structure makes Gemini 3.7 Flash particularly relevant to applications that generate large numbers of model calls, including autonomous agents and developer tools.

Availability for Developers

Developers can access Gemini 3.7 Flash through several Google platforms.

Google lists availability through:

  • Google Antigravity

  • Gemini API

  • Google AI Studio

  • Android Studio

Google also provides a developer guide for integrating the model into applications.

For enterprise customers, Gemini 3.7 Flash is available through the Gemini Enterprise Agent Platform and the Gemini Enterprise application.