Google Launches Gemini 3.7 Flash with Coding and Agent Improvements

My take: Google launched Gemini 3.7 Flash three weeks after the previous version, and the coding numbers are relevant: it climbs from 34.4% to 43.6% on FrontierCode 1.1 and from 49.0% to 65.3% on DeepSWE v1.1. Worth noting that these figures are reported by Google on its own model, so it is smart to read them with your own judgment and wait for independent validation before taking them as definitive.

The pricing is also worth attention: $0.75 per million input tokens and $3.75 output until December 31, 2026. For developers and creators working with code, agents, or document-heavy tasks, that combination of price and coding performance is a concrete proposition worth testing.

And yes, the flagship Gemini model (Gemini 3.5 Pro) is still delayed. Instead of the expected flagship, frequent Flash updates arrive, which is the model most people run in production anyway. The practical question: do you already have a process to evaluate new models as soon as they drop, before your competition integrates them first?