Google unveils Gemini 3.7 Flash with enhanced coding and agentic capabilities

Google has announced Gemini 3.7 Flash, just three weeks after Gemini 3.6 Flash, positioning it as a model for developers and businesses using multi-step AI systems, especially in software development and process automation. According to Google, the new model shows notable gains in debugging, problem-solving, and production code generation, with benchmark scores on FrontierCode 1.1 Main rising from 34.4% to 43.6% and on DeepSWE v1.1 from 49% to 65.3%, though these figures come from Google's own evaluations. The model also improves agentic capabilities, helping to reduce human intervention by planning actions, using tools, and adapting to obstacles. In web development, Gemini 3.7 Flash reportedly achieves an Elo score of 1588 on the WebDev Arena, compared to 1538 for the previous version, and shows progress in complex document analysis in fields like finance, law, and biosciences. The model is also integrated into Gemini Spark, Google's continuous personal agent for Google Workspace users, available to subscribers of Google AI Pro and Ultra in over 160 countries. Pricing is set at $0.75 per million input tokens and $3.75 per million output tokens until the end of 2026, rising to $1.50 and $7.50 respectively from January 1, 2027.