Google has launched Gemini 3.7 Flash, its latest artificial intelligence model designed specifically for software coding and automated business tasks, according to an announcement. Released three weeks after Gemini 3.6 Flash, the new model is being pitched as a lower-cost option for businesses building autonomous AI systems capable of planning tasks, utilizing software tools, and completing multi-step workflows with reduced human intervention.
Google Launches Gemini 3.7 Flash for Coding and Business Automation
To drive adoption, Google is offering Gemini 3.7 Flash at an introductory rate of 75 cents per million input tokens and $3.75 per million output tokens through the end of the year, which is half the original cost of the 3.6 Flash version. Additionally, the model is rolling out immediately to Reuters, the company’s subscription-based AI agent service available to Google AI Pro and Ultra customers across more than 160 countries. The model has also been made available in AI Mode for Search, as noted in updates from Theverge.
Performance Gains in Coding and Web Development
According to a Google blog post cited in reports, Gemini 3.7 Flash demonstrates strong gains over its predecessor in coding tasks such as debugging and issue resolution. It achieves higher first-pass code accuracy and improved performance in generating production-ready code. Benchmarks highlight these improvements, showing scores of 43.6% versus 34.4% in FrontierCode 1.1 Main, and 65.3% versus 49.0% in DeepSWE v1.1.
In web development, the model generates more functional layouts and feature-complete apps using fewer prompts. For user interface generation, it shows high design adherence and parity based on reference inputs like screenshots, images, or full design systems. Furthermore, it outperforms 3.6 Flash on Arena.ai’s WebDev Arena, registering an Elo score of 1588 compared to 1538.
Reasoning Improvements in Knowledge-Dense Fields
For knowledge-dense fields including finance, law, and biosciences, Gemini 3.7 Flash delivers enhanced reasoning and accuracy. It significantly outperforms Gemini 3.6 Flash on the GDP.pdf benchmark—an evaluation testing a model’s ability to process complex documents—scoring 34.0% against 22.0%.
The model also surpasses its predecessor in AutomationBench, demonstrating an increased capability to effectively complete real-world business workflows with a score of 30.4% compared to 17.0%.
Broader Context and Leadership Changes at Google
The launch accompanies a period of transition and intense competition within Google’s parent company, Alphabet (GOOGL.O), opens new tab. Investors have closely watched for developments regarding Google’s premium model, Gemini 3.5 Pro, as a test of whether its DeepMind AI unit can keep pace with rivals Anthropic and OpenAI. Google previously stated in July that Gemini 3.5 Pro was being tested with partners and would arrive soon, but officials offered no details on when the flagship Pro model will be released alongside the 3.7 Flash launch.
Amid these strategic pushes, co-founder Sergey Brin has urged key AI staff to go all-in on the Gemini model. Concurrently, the tech giant announced a sweeping leadership overhaul of its Google DeepMind AI division, during which chief Demis Hassabis stepped aside in favor of his deputy, Koray Kavukcuoglu, while two original technical co-leads of Gemini departed to co-found a startup. CEO Sundar Pichai previously defended the company’s AI strategy during a July earnings call, pushing back on concerns that delays in flagship models had caused Google to cede ground in AI coding.
