Google introduced Gemini 3.7 Flash on August 13, 2026. The company calls it the most intelligent workhorse yet in the Flash line for coding and agents. It arrived three weeks after Gemini 3.6 Flash.
What Google actually claimed
Google says 3.7 Flash is better at software engineering, knowledge work, and web development than 3.6 Flash. On FrontierCode 1.1 Main it reports 43.6 percent versus 34.4 percent. On DeepSWE v1.1 it reports 65.3 percent versus 49.0 percent. On Arena.ai's WebDev Arena it reports an Elo of 1588 versus 1538. On GDP.pdf, a document-comprehension test, it reports 34.0 percent versus 22.0 percent. On Zapier's AutomationBench it reports 30.4 percent versus 17.0 percent.
Those are Google's published numbers, not independent Alberta shop results. Treat them as a buying signal, not as a promise about your tender file or your CRM.
The price is the other headline
Introductory pricing is $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. On January 1, 2027, Google says that rises to $1.50 and $7.50. If a team builds a daily agent on the sale price, the January bill can jump.
Developers can try it in Google Antigravity, the Gemini API, Google AI Studio, and Android Studio. Enterprises can reach it through Gemini Enterprise. Google AI Pro and Ultra subscribers in supported countries get it inside Spark. Spark is not listed for the European Economic Area, Nigeria, Switzerland, or the United Kingdom.
What a local operator should do
Do not rip out Claude, Grok, or Codex because a new Flash number dropped. Pick one coding or document job, keep a human review step, and compare error rate and cost against the model you already pay for. If the work touches customer files, quotes, or legal language, keep it off public prompts.