Gemini 3.7 Flash Blazes New Trails for Google's AI Models

Google has released Gemini 3.7 Flash, its latest iteration of the Gemini family of models. The new model boasts significant improvements over its predecessor, Gemini 3.6 Flash, with a 4-point increase in Intelligence Index scores.

Gemini 3.7 Flash (high) achieves an average Time per Task of 1.7 minutes, making it 40% faster than GPT-5.6 Terra (max). This puts Gemini 3.7 Flash on the Intelligence vs. Time per Task Pareto frontier, highlighting Google's focus on speed.

The new model also excels in agentic evaluations, including Tau3 Banking (+3 points), Terminal-Bench v2.1 (+8 points), and GDPval-AA v2 (+103 Elo). Gemini 3.7 Flash (high) costs $0.40 per Intelligence Index task at the discounted price of $0.75/$3.75 per 1M tokens, a 30% reduction from Gemini 3.6 Flash.

Gemini 3.7 Flash leads on AutomationBench-AA and AA-AnalystAgent benchmarks, demonstrating its agentic capabilities in simulated SaaS environments. The model's multimodality features text, image, video, and speech input with text output, retaining the same context window as Gemini 3.6 Flash.