Google introduces Gemini 3.7 Flash. The model scores significantly higher on benchmarks for coding and agentic AI than its predecessor, 3.6 Flash, and is temporarily half the price: $0.75 per million input tokens.
Gemini 3.7 Flash is released just three weeks after 3.6 Flash. Google calls the model its most intelligent workhorse to date for coding and agents, and cites developer feedback and algorithmic improvements as the direct reason for the rapid follow-up.
Google is making the biggest strides in software tasks. FrontierCode 1.1 Main, for generating production-ready code, goes from 34.4% to 43.6%, while DeepSWE v1.1, for practical program quality, goes from 49.0% to 65.3%. In web development, according to Google, the model delivers more functional layouts with fewer prompts; on Arena.ai’s WebDev Arena, the Elo score rises from 1,538 to 1,588.
Knowledge-intensive work is also receiving attention. For finance, law, and the life sciences, Google claims improved reasoning capabilities. On GDP.pdf, a benchmark for processing complex documents, the score rises from 22.0 to 34.0 percent. AutomationBench, which tests real-world business workflows, jumps from 17.0 to 30.4 percent.
Price halved through the end of this year
Through the end of this year, Google is charging an introductory price of $0.75 per million input tokens and $3.75 per million output tokens. After that, the rates will return to the regular prices of $1.50 and $7.50, respectively. By comparison, Gemini 3 Flash is priced at $0.50 and $3.00, making it still the most affordable option in the series.
Beyond the numbers, Google emphasizes behavioral improvements. The model is said to handle obstacles better, ask for clarification more often, and follow instructions more accurately. Google states: “A more disciplined execution means less manual oversight and fewer retries across engineering workflows.”
Gemini Spark is now available on 3.7 Flash. Google also provides the model with updated Frontier Safety safeguards against misuse related to CBRN and cyber offense. Developers can access it via Google Antigravity, the Gemini API through AI Studio, and Android Studio. Business users can find 3.7 Flash in the Gemini Enterprise Agent Platform and the Gemini Enterprise app.