Google AI Releases Gemini 3.7 Flash for Coding and Agents

Google AI released Gemini 3.7 Flash, the latest iteration in its Flash model tier, approximately three weeks after the introduction of Gemini 3.6 Flash. This new model is described in its model card as a refinement of Gemini 3.6 Flash, incorporating algorithmic enhancements to its core reasoning capabilities rather than undergoing a new pretraining process. Gemini 3.7 Flash supports multimodal inputs, accepting text, images, audio, and video, and features an extensive 1 million token context window. It can generate output tokens up to 64,000 and offers customizable thinking configurations that allow users to balance output quality against cost and latency. The model's knowledge cutoff remains March 2026. Significant improvements are noted in three key areas: software engineering, document-intensive knowledge work, and web development. A primary focus of this release is its pricing structure. Gemini 3.7 Flash is priced at $0.75 per 1 million input tokens and $3.75 per 1 million output tokens. This represents a 50% reduction compared to the initial pricing of Gemini 3.6 Flash and is approximately one-third the blended cost of competing models such as Claude Sonnet 5 or GPT-5.6 Terra. Access to Gemini 3.7 Flash is exclusively available through API and enterprise channels, with no open-weight versions provided. Users can interact with the model via hosted platforms including the Gemini API, Google AI Studio, Google Antigravity, Android Studio, the Gemini Enterprise Agent Platform, and the Gemini Enterprise app. Consumer access is facilitated through Gemini Spark, available on Google AI Pro and Ultra plans. The pricing strategy is particularly beneficial for startups and mid-market companies, making the deployment of always-on agents economically feasible without requiring a Pro-tier budget. Regulated enterprises can leverage Gemini Enterprise for a governed access path. However, organizations with strict data residency requirements or those operating in air-gapped environments will not be able to utilize this model, as self-hosting is not an option. Google's internal evaluations suggest suitability for industries such as legal services, financial services, biosciences, and enterprise operations, with performance highlighted in benchmarks like Harvey LAB-AA, GDP.pdf, and AutomationBench. Specific applications include the development of long-running coding agents, automation of document-heavy back-office processes, generation of user interfaces from screenshots or design systems, and the creation of structured data pipelines from PDF documents. In terms of performance benchmarks, Gemini 3.7 Flash achieved a score of 43.6% on the FrontierCode 1.1 Main benchmark, which evaluates production code quality. This represents an improvement over the 34.4% score achieved by Gemini 3.6 Flash on the same benchmark.