DeepMind introduces Gemini 3.7 Flash for coding and agents
DeepMind announced Gemini 3.7 Flash on Aug. 13, 2026, describing it as the most capable workhorse model in the Flash series for coding and agent workflows. The release follows Gemini 3.6 Flash by three weeks and is presented as a response to developer feedback and algorithmic improvements.
According to the company, the model improves debugging, issue resolution and first-pass code accuracy. Its results on FrontierCode 1.1 Main and DeepSWE v1.1 were 43.6% and 65.3%, compared with 34.4% and 49.0% for Gemini 3.6 Flash. It also scored 1588 versus 1538 on Arena.ai’s WebDev Arena. On GDP.pdf, performance was 34.0% versus 22.0%, while AutomationBench results were 30.4% versus 17.0%.
Until the end of the year, introductory pricing is $0.75 per 1M input tokens and $3.75 per 1M output tokens. Gemini Spark will use the new model for Google AI Pro and Ultra subscribers in more than 160 countries. DeepMind also said Gemini 3.7 Flash includes updated safeguards for CBRN misuse and cyber offense.
Written by Sakhanda Wire AI from reporting by DeepMind. Facts, quotes and full context are in the original article below.