Google has released Gemini 3.7 Flash, a new AI model with significant advancements in programming and agent tasks. The introductory price is $0.75 per million input tokens, half of what it was at the launch of its predecessor Gemini 3.6 Flash. Users in the EU, UK, and Switzerland are currently excluded from the new AI agent app Gemini Spark with the model.
Benchmark values significantly increase for programming tasks
Google released Gemini 3.7 Flash on August 13, 2026, as the company states in the official blog – just three weeks after Gemini 3.6 Flash. The value in the programming benchmark FrontierCode rises from 34.4 to 43.6 percent, and for lengthy software engineering tasks in the DeepSWE test from 49.0 to 65.3 percent. The model also improves in web development (WebDev Arena, Elo rating from 1538 to 1588) and in understanding complex PDF documents (GDP.pdf benchmark, from 22.0 to 34.0 percent). The model adapts better to obstacles, clarifies intentions when needed, and follows instructions more precisely than its predecessor. At the same time, the company has updated the safeguards against misuse for chemical, biological, radiological, and nuclear purposes as well as for cyberattacks.
Introductory price halved by the end of the year
Until December 31, 2026, Gemini 3.7 Flash costs $0.75 per million input tokens and $3.75 per million output tokens. Starting January 1, 2027, both values will double to $1.50 and $7.50, respectively. The introductory price thus undercuts the launch price of Gemini 3.6 Flash by half. In the automation of business processes (AutomationBench), the value rises from 17.0 to 30.4 percent – an area that particularly responds to the price per token with a high volume of recurring tasks. With this step, Google joins an industry-wide price competition: Just in July, OpenAI reduced the prices for its model GPT-5.6 by up to 80 percent.
Gemini Spark remains blocked for the EU for now
The app Gemini Spark, Google’s AI agent for end customers, now runs with Gemini 3.7 Flash and is available in more than 160 countries for subscribers of Google AI Pro and Ultra. The European Economic Area, the UK, Switzerland, and Nigeria remain excluded. The specialist portal PPC Land suspects regulatory requirements such as the EU AI Act and European data protection as the reason for the exclusion; Google does not provide an official explanation. Developers and companies in Germany can still access the new model: through Google AI Studio, the Gemini API, and the platforms Vertex AI and Gemini Enterprise, Gemini 3.7 Flash is regularly available. Thus, only the new, independently operating consumer app is affected, not access for professional applications. The exclusion fits into an already known gap: Google’s flagship model Gemini 3.5 Pro is further delayed due to weak internal programming tests, and a new date has not been set.
It will be crucial whether Google lifts the app ban for Spark in 2026 or permanently refers European users to the developer tools. For IT departments in Germany, the current situation means that while they can utilize the lower prices and better benchmark values through the programming interface, they are currently not allowed to offer the finished consumer product to their customers.
