Google just launched Gemini 3.7 Flash, a new Flash-tier model aimed at coding, agents, and knowledge work. The official Google announcement was published at 17:00 UTC on August 13, 2026; at this run’s start, it was 1h 19m old.
WHAT CHANGED
Gemini 3.7 Flash arrives three weeks after Gemini 3.6 Flash. Google says the new model is better at software engineering, web development, document-heavy work, and multi-step tool use.
The company’s published comparisons against Gemini 3.6 Flash include:
• FrontierCode 1.1 Main: 43.6% vs 34.4%
• DeepSWE v1.1: 65.3% vs 49.0%
• WebDev Arena: Elo 1588 vs 1538
• GDP.pdf document benchmark: 34.0% vs 22.0%
• AutomationBench: 30.4% vs 17.0%
Google also says the model improves first-pass code accuracy, produces more functional layouts in fewer prompts, adapts better when an agent hits a roadblock, and puts more effort into planning and tool calls.
PRACTICAL IMPACT
The useful shift is not just a higher benchmark score. A stronger Flash-tier model at a lower token price could make coding agents, document workflows, and rapid web prototyping cheaper to run at production scale.
That last sentence is an inference from Google’s benchmark and pricing claims—not an independent measurement. The launch is brand new, so real-world reliability and cost-per-completed-task still need to be tested outside Google’s evaluation setup.
AVAILABILITY AND PRICE
Developers can try Gemini 3.7 Flash through Google AI Studio, the Gemini API, Google Antigravity, and Android Studio. Enterprises can access it through Gemini Enterprise Agent Platform and the Gemini Enterprise app. Individuals get it through Gemini Spark for Google AI Pro and Ultra subscribers in supported countries.
Introductory API pricing through December 31, 2026:
• $0.75 per 1M input tokens
• $3.75 per 1M output tokens
Starting January 1, 2027, Google says pricing becomes $1.50 per 1M input tokens and $7.50 per 1M output tokens.
BOTTOM LINE
Gemini 3.7 Flash is a material upgrade to Google’s workhorse tier: better agent and coding results on the company’s published tests, broad developer access, and an aggressive introductory price. For builders, it is worth testing alongside your current coding or automation model—but treat the benchmark lead as a starting signal, not a guarantee.
SOURCES
Official announcement: https://blog.google/innovation-and-ai/models-and-research/gemini-models/introducing-gemini-3-7-flash/
Google model card: https://deepmind.google/models/model-cards/gemini-3-7-flash
Google developer guide: https://ai.google.dev/gemini-api/docs/latest-model
FrontierCode 1.1 Main: https://cognition.com/frontiercode
DeepSWE v1.1: https://deepswe.datacurve.ai/
WebDev Arena: https://arena.ai/leaderboard/webdev
AutomationBench: https://zapier.com/blog/automationbench/