Google released an updated version of its entry-level artificial intelligence model today, promising faster coding assistance and better performance on enterprise tasks.
New model outperforms rivals in coding, document analysis
Gemini 3.7 Flash arrives three weeks after its predecessor. Despite the short development cycle, Google says the model outperformed comparable offerings from Anthropic and OpenAI across nine benchmarks.
One of the tests where it took first place was FrontierCode 1.1 Main. The benchmark includes 100 programming tasks in multiple languages. Models must generate working code, follow project-specific style guides, and pass bug testing before submission.
Related: Skan AI Raises 63 Million Dollars
The new model also shows improved ability to generate user interfaces. According to Google, layouts designed by Gemini 3.7 Flash more closely match reference images uploaded by users. It can process up to 1 million tokens worth of images, video, and text per prompt, with responses reaching 64,000 tokens.
“Gemini 3.7 Flash delivers a noticeably improved developer experience over 3.6 Flash,” said Tulsee Doshi, Google’s senior director of product management. “It better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity. It thinks more diligently, putting in more effort into multi-step planning and tool calls.”
Beyond coding, the model performed well on business document analysis. In the GDP.pdf benchmark, which tests a model’s ability to answer questions about corporate paperwork, Gemini 3.7 Flash answered 34% of questions correctly. That put it 6% ahead of Anthropic’s Claude Sonnet 5 and 9.3% ahead of GPT-5.6 Terra.
Related: AI solves long open math problems
Google also says the model can power AI agent ensembles.
Pricing cut in half, consumer integration coming
The company didn’t disclose the model’s architecture. Its model card indicates it’s based on the same design as Gemini 3.6 Flash, which itself derived from Gemini 3 Pro—a transformer-based mixture-of-experts system. Training details were also omitted.
Developers will be able to access Gemini 3.7 Flash for half the price of the previous version through the end of the year. The model will also be integrated into Gemini Spark, a consumer-focused AI agent that can browse the web and interact with other Google services. Gemini Spark debuted in March.