Gemini 3.7 Flash: the surprising leap and the temporary price drop from Google

Gemini 3.7 Flash represents a significant advance over Gemini 3.6 Flash. This new version improves especially in programming, automation, and understanding complex PDF documents, as demonstrated in various official Google technical tests. The performance increase is due to the model’s better ability to follow multiple instructions and plan chained steps, a key necessity for users working with automated processes or coding.

The source of this information comes from Google and official developer documents testing this model since August 13, 2026, on the Spark platform for AI Pro and Ultra subscriptions.

Key technical improvements of Gemini 3.7 Flash

What are the most notable advances?

Gemini 3.7 Flash nearly triples performance in the Terminal-bench 3.0 test, increasing from 5.4% to 14.9%. This test measures how the model works with a terminal, highly relevant for developers and system administrators. Additionally, in AutomationBench, which evaluates automated workflows, the result almost doubles: from 17% to 30.4%. In programming, improvements are clearly visible, with an increase from 48.6% to 65.3% in DeepSWE v1.1 and from 34.4% to 43.6% in FrontierCode 1.1.

Why do these improvements matter to users?

For anyone using Gemini to program or manage automated processes, this means fewer errors and fewer attempts to get functional code or workflows. This translates into saved time and less frustration. Furthermore, the 55% improvement in interpreting complex PDF documents (from 22% to 34%) broadens the possibilities of use in professional environments that require handling extensive or technical documents.

How to make the most of the new Version and the price reduction

How to activate Gemini 3.7 Flash on Spark?

Users with AI Pro and Ultra subscriptions already have automatic access to Gemini 3.7 Flash since August 13, 2026. If you have one of these subscriptions, the change has been applied without you needing to do anything. Developers using the Google API can continue using Gemini 3.6 Flash, as both versions are available.

What is the current Rate and what will happen to the price?

Until December 31, 2026, Gemini 3.7 Flash can be used at a promotional price of $0.75 per million input tokens and $3.75 per million output tokens. This rate represents half the initial cost of Gemini 3.6 Flash. But beware: starting January 1, 2027, the price will double to $1.50 and $7.50 respectively, unless there are changes in Google's policy.

How to check if you are taking advantage of the Improvement and the promotional Rate

How to know if you are using Gemini 3.7 Flash?

If you are a Spark user with an AI Pro or Ultra subscription and have been using the model since August 13, you are. If you use the API and see you are working with Gemini 3.6 Flash, you can choose to upgrade when possible.

What signs confirm the model is working correctly?

• Better accuracy in code and fewer corrections needed.
• Smoother automation flows without unexpected errors.
• More accurate interpretation of complex PDF documents.
• Improved responses to queries requiring multiple steps or planning.

To ensure you benefit from Gemini 3.7 Flash and the promotional rate, periodically review your account settings and subscription. Google will announce any changes in advance.

We also recommend enabling two-factor authentication and reviewing access permissions to protect data processed by the AI. Making backups and staying alert for possible anomalies in model responses will prevent long-term issues.

This release of Gemini 3.7 Flash demonstrates that Google is strongly committed to improving productivity for professional users with more effective and accessible artificial intelligence, but the temporary price reduction requires planning for future use and costs. It is an opportunity worth taking advantage of right now.