Google brings Gemini 3.7 Flash just three weeks after the previous model's release. It entices with better code at half the price

All the news from Google related to Gemini artificial intelligence has been difficult for a layman to follow even until now. The constantly expanding range of services doesn’t help, nor does the pace of introducing the latest models contribute to clarity. The cheaper and faster Flash series now gets another addition just a few weeks after the introduction of the previous one – and Gemini 3.7 Flash is expected to be smarter mainly in coding, moreover at roughly half the price of its predecessor.

Gemini 3.7 Flash will mainly appeal to developers

Google positions Gemini 3.7 Flash as its smartest all-around model yet for coding and so-called agent tasks – situations where the model plans its own steps and calls various tools to execute them. The company claims that the new version was developed directly based on feedback from developers who worked with the previous 3.6 generation.

The progress is expected to be most evident in programming. In the DeepSWE v1.1 test, which measures how well a model handles a real software bug, Google reports a success rate of 65.3% compared to 49% for version 3.6 Flash. In the more demanding FrontierCode 1.1 benchmark, the newer model completed 43.6% of tasks compared to the previous 34.4%. The company also claims a similar lead in website creation or in reading complex documents from the fields of finance and law.

Beyond dry charts, Google also promises more pleasant collaboration. The model reportedly better recognizes when to ask for clarification on a task, adheres more consistently to instructions, and for more complex tasks, according to the company, thinks more ahead before tackling them. In practice, this should mean less manual oversight and fewer repeated attempts before the agent completes the task.

The price is also attractive

Perhaps even more appealing than performance, however, is the price tag. Gemini 3.7 Flash starts at an introductory price of $0.75 per million input tokens and $3.75 per million output tokens – roughly half of what 3.6 Flash originally cost. Tokens can be simply imagined as pieces of text that the model reads as input and generates as output; the more that flow through, the more you pay.

Moreover, this introductory rate is only valid until the end of the year, after which the price will likely increase. And Google is far from finished – the company’s CEO recently hinted at a significant leap in the form of Gemini 4.

Google brings Gemini 3.7 Flash just three weeks after the previous model's release. It entices with better code at half the price Adam Kurfürst News Adam Kurfürst News

Availability can be problematic, especially in Europe

For developers, the model is available almost immediately – via the Gemini API in Google AI Studio, in Android Studio, and in the new Antigravity environment. However, it’s harder for the average user. The new version primarily powers the personal agent Gemini Spark, which is supposed to continuously handle tasks for you, but the European Economic Area, and thus the Czech Republic, remains excluded from its offering.

Czechs will therefore not be able to use Spark with Gemini 3.7 Flash yet, and will only be able to access the new model indirectly through developer tools instead of simply working with it directly in the Gemini application. Google is also adding security safeguards to the model to make it harder to misuse, for example, in cyberattacks.

Are you trying out the new Gemini models for programming, or are the larger ones sufficient for you for now?

Source: Google