Google Introduces Gemini 3.7 Flash, the Smartest Model for Coding and Agent Tasks

Summary

Google announced the Gemini 3.7 Flash model and positions it as the smartest core AI for coding and agent tasks; it delivers performance gains over the 3.6 Flash version.

Why it matters

The model's enhanced coding abilities and larger context window (up to 1M tokens) can accelerate complex development workflows. Its temporary half-price pricing could lower entry barriers for developers and businesses, potentially speeding up AI adoption in software engineering.

Highlights

• Achieves higher scores on coding benchmarks (FrontierCode 43.6% vs 34.4%, DeepSWE 65.3% vs 49.0%).
• Offers a million-token context window and output support up to 64,000 tokens.
• Provides an introductory price of $0.75 per 1M input tokens and $3.75 per 1M output tokens for a limited time, which is half the cost of Gemini 3.6 Flash.

Critical take

The performance claims rely on Google's internal benchmarks without third-party verification, and the price cut is temporary; long-term costs will revert to standard rates and may still be high compared to competing models.