Google just rolled out Gemini 3.7 Flash, its latest AI model in the Flash series, and it’s built for coding, AI agents, web development, and heavy-duty knowledge work. The launch comes only three weeks after Gemini 3.6 Flash, but this new version steps things up with sharper coding accuracy, better multi-step reasoning, and smoother tool integration. Plus, Google slashed the introductory price to half what they charged for Gemini 3.6.
Gemini 3.7 Flash focuses on complex AI workflows
Gemini 3.7 Flash is all about complex AI workflows. Google calls it their smartest workhorse yet for coding and agent-driven tasks. They tuned the model to handle detailed, multi-layer projects, push through tricky barriers, and follow developer instructions more closely. It's designed so developers don’t have to step in as often or run multiple retries, especially when tackling big engineering jobs or tool-heavy processes.
Major coding improvements
On the coding front, Gemini 3.7 Flash beats 3.6 Flash in debugging, fixing issues, and spitting out production-ready code. On FrontierCode 1.1 Main, Gemini 3.7 Flash scored 43.6 per cent—up from 34.4 per cent for its predecessor. DeepSWE v1.1 numbers are even better: 65.3 per cent versus the previous model’s 49 per cent. Google points out the model’s first-pass code generation is more reliable, which means developers spend less time cleaning up.
Better at building websites and apps
For web development, the improvements keep coming. Gemini 3.7 Flash builds more functional layouts and full-featured apps using fewer prompts. It can handle visual references—screenshots, images, design systems—and keep everything consistent.
In Arena.ai’s WebDev Arena, Gemini 3.7 Flash earned an Elo score of 1588, while Gemini 3.6 Flash managed 1538.
Stronger performance for knowledge work
When it comes to knowledge work, especially in fields like finance, law, and biosciences, Google says Gemini 3.7 Flash really delivers. On the GDP.pdf benchmark for reading complex documents, it scored 34 per cent, up from Gemini 3.6 Flash’s 22 per cent. On AutomationBench, it hit 30.4 per cent—almost twice the previous model’s 17 per cent.
Gemini Spark gets the new model
Google is further putting Gemini 3.7 Flash into Gemini Spark- its personal AI agent for Google AI Pro and Ultra subscribers across more than 160 countries. This upgrade boosts Spark’s tool use in Google Workspace apps and raises the bar for accuracy and workflow quality.
Gemini 3.7 Flash costs half as much
As for price, Gemini 3.7 Flash is a bargain compared to the last model. Through the end of the year, Google’s charging just USD 0.75 per 1 million input tokens and USD 3.75 per 1 million output tokens. Developers can tap into the model via Google AI Studio, Gemini API, Google Antigravity, and Android Studio. For businesses, it’s available through the Gemini Enterprise Agent Platform and the Gemini Enterprise app.
ALSO READ:
Google Assistant to be discontinued from September 4, as Gemini set to take over: Benefits here
Gemini to become default AI assistant on Android from September 2026: Google Assistant to shut down
Google launches Gemini 3.6 Flash and Gemini 3.5 Flash-Lite: Gemini 4 already in training
