It seems the race in the world of artificial intelligence has shifted into high gear; just three weeks after unveiling the Gemini 3.6 Flash model, Google has surprised everyone by launching the new Gemini 3.7 Flash. Google has specifically designed this upgraded model to handle complex programming tasks, build AI agents, and manage professional workflows, while offering tangible improvements in speed, accuracy, and overall cost reduction for developers.
A Huge Leap in Software Writing and Web Development
The new model has shown remarkable superiority in programming performance benchmarks and code handling, with the success rate in the FrontierCode 1.1 Main test rising from 34.4% to 43.6%, while achieving a significant jump in the DeepSWE v1.1 test from 49.0% to 65.3%. Google confirms that the model now generates code that is more ready for direct use in production environments, with increased first-attempt accuracy and reduced programming errors.
As for web application development, Gemini 3.7 Flash has demonstrated higher efficiency in creating fully-featured interfaces and layouts using fewer prompts and text instructions. This is reflected in the model’s increased rating on the WebDev Arena Elo index from 1538 to 1588 points, making it a practical and reliable choice for web designers and developers.
Deeper Understanding of Documents and Autonomous Intelligent Work
The upgrade is not limited to writing code; it extends to processing large documents and intelligently extracting data from them. The model’s performance in the GDP.pdf test, dedicated to document analysis, rose from 22.0% to 34.0%, while it recorded an excellent climb in the AutomationBench test from 17.0% to 30.4%.
Gemini 3.7 Flash adapts better when encountering obstacles during task execution, as it inquires about ambiguous details when needed and adheres to instructions with extreme precision. This reduces manual human intervention and saves time for programming and engineering teams that rely on AI to manage their projects.
Gemini Spark Upgrade, Pricing, and Availability
In parallel with this launch, Google has begun transitioning the Gemini Spark AI agent to run entirely on the new 3.7 Flash model. This update is now available to Google AI Pro and Ultra subscribers in over 160 countries, providing them with better skills in integrating with Google Workspace applications, such as drafting emails, summarizing files, and updating daily documents quickly and smoothly.
Regarding cost, Google has reduced the introductory usage prices by 50% compared to the initial launch of 3.6 Flash; the cost is currently $0.75 per million input tokens and $3.75 per million output tokens until the end of this year.
Developers can access the model via the Google Antigravity platform, the Gemini API interface in Google AI Studio, and Android Studio, in addition to platforms dedicated to institutions and large companies.
Google has bolstered the new model with advanced security and protection packages within the Frontier Safety framework to mitigate risks and dangers related to cyberattacks or misuse in sensitive areas.
Source:
Leave a Reply