Google released Gemini 3.7 Flash three weeks after Gemini 3.6 Flash, marking the third iteration of its workhorse coding and agent model line. The new model shows substantial gains in software engineering and knowledge work: FrontierCode 1.1 Main scores jump from 34.4% to 43.6%, and DeepSWE v1.1 goes from 49.0% to 65.3%. For web development, the Elo score on Arena.ai's WebDev Arena climbs from 1538 to 1588, enabling feature-complete app generation with fewer prompts.
Gemini 3.7 Flash ships at an introductory price of $0.75/1M input and $3.75/1M output tokens—half the original 3.6 Flash pricing at launch. The model demonstrates improved reasoning on knowledge-dense domains (finance, law, biosciences), benchmarking 34.0% on GDP.pdf (up from 22.0%) and 30.4% on AutomationBench for real-world business workflows (up from 17.0%). Developer experience improves with better multi-step planning and fewer retries in engineering workflows.
Gemini 3.7 Flash integrates into Gemini Spark (Google's 24/7 personal agent for Pro/Ultra subscribers), Google Antigravity, Gemini Enterprise Agent Platform, and the Android Studio dev experience. Google's Flash line—now releasing at ~3-week cadence—serves as the company's high-frequency innovation engine for agentic and coding workloads, competing directly with OpenAI's tiered GPT-5.6 and Anthropic's Claude family on speed and cost.