Google on Wednesday released Gemini 3.7 Flash, a coding- and agent-tuned update to its small workhorse model, and priced the new release at half the cost of the Gemini 3.6 Flash version it introduced just three weeks earlier.
The launch keeps Google on a rapid release cadence for its economy-tier models even as its flagship Gemini 3.5 Pro, first promised for July, remains unreleased. Through year-end, developers can use the new model for 75 cents per million input tokens and $3.75 per million output tokens, a 50 percent introductory discount from 3.6 Flash prices, Reuters reported.
Where it ships
The model is deploying to Gemini Spark, Google's subscription service for Pro and Ultra tier customers across more than 160 countries, according to Reuters. It is also rolling out inside GitHub Copilot's Pro, Pro+, Max, Business and Enterprise tiers, covering Visual Studio Code, Visual Studio, Copilot CLI, the GitHub Copilot cloud agent, the GitHub Copilot app, JetBrains, Xcode and Eclipse, Axios reported.
The benchmarks
Google said the model outperformed comparable offerings from Anthropic and OpenAI across nine benchmarks, according to SiliconANGLE. It reached first place on FrontierCode 1.1 Main, a 100-task programming test that grades both code functionality and compliance with enterprise style and bug-testing rules. On GDP.pdf, a business-document comprehension test, Gemini 3.7 Flash answered 34 percent of questions correctly, ahead of Claude Sonnet 5 by 6 percent and GPT-5.6 Terra by 9.3 percent. The model can process 1 million tokens of combined text, images and video per prompt and return responses of up to 64,000 tokens.
Tulsee Doshi, Google's senior director of product management, said the model "better adapts to roadblocks, clarifies intent when needed, and follows instructions with greater fidelity."
What is missing
Anthropic and OpenAI had not published counter-benchmarks against Gemini 3.7 Flash by press time, and Wednesday's reporting from Reuters, SiliconANGLE and Axios drew on Google's own scores against unnamed prior versions of the rival systems. Google gave no firm date for the delayed Gemini 3.5 Pro flagship, saying it would arrive soon. Reuters noted that recent leadership changes at Google DeepMind have accompanied the company's push to keep pace with OpenAI and Anthropic, both of which have shipped multiple mid-year model updates.
The next test will be that flagship launch, which Axios has cast as a bellwether for whether Google's foundation-model roadmap can keep pace.