Google Unveils Gemini 3.7 Flash, Touting Improved Performance

Google has released the Gemini 3.7 Flash model, just three weeks after its predecessor, Gemini 3.6 Flash. According to Senior Director Tulsee Doshi, this new model boasts improved coding and agentic performance, with notable gains in benchmarks such as FrontierCode 1.1 Main test (34.4% to 43.6%) and DeepSWE v1.1 (49% to 65.3%). The WebDev Arena score has also risen from 1,538 to 1,588.

The new model is positioned as a workhorse, designed to counter the lower costs of competing models with its own introductory price. Doshi highlights the performance improvements in processing complex documents (GDP.pdf benchmark up 34% from 22%) and executing common business workflows (AutomationBench score rose to 30.4% from 17%).

However, some may question the need for a new model release so soon after the previous one, with critics suggesting that this may be more about maintaining the appearance of constant improvements in Google's AI.