Google unveiled Gemini 3.7 Flash in August 2026, a specialized model for complex software engineering and enterprise agentic workflows that features a 50% price cut per token.
The model achieved a 43.6% score on the FrontierCode 1.1 benchmark and 65.3% on DeepSWE v1.1, demonstrating significant improvements in debugging and production-ready code generation over its predecessor.
Pricing is set at USD 0.75 per million input tokens and USD 3.75 per million output tokens until 31 December 2026, with access provided via Google’s Gemini API and AI Studio.