Just three weeks after the release of Gemini 3.6 Flash, Google has released the new Gemini 3.7 Flash model. The company states that this rapid development process was achieved thanks to developer feedback and new algorithmic innovations.

The new model promises significant performance improvements in software development, web design, and information-intensive work areas. Significant gains are said to have been achieved, particularly in debugging and problem-solving processes, compared to the previous version.

Enhanced Coding and Analysis Capabilities

Gemini 3.7 Flash has increased its score in the DeepSWE v1.1 benchmark test in software engineering from 49.0% to 65.3%. In the FrontierCode 1.1 Main test, the success rate increases from 34.4% to 43.6%.

On the web development side, the model can create more functional layouts and complete applications with fewer commands. In Arena.ai WebDev Arena tests, the Elo score, which was 1538, reaches 1588 with the new version.

In interface creation processes, the model shows high fidelity to reference inputs such as screenshots or design systems. In fields such as finance, law, and bioscience, significant improvements are noticeable in processing complex documents.

In the GDP.pdf benchmark test, the model achieves a 34.0% success rate, surpassing the previous version’s performance of 22.0%. In the AutomationBench test, the success rate in completing real-world workflows increased from 17.0% to 30.4%.

It is stated that the model works more carefully in multi-step planning and tool usage. This disciplined work structure requires less manual intervention and fewer repetitions in engineering workflows.

Security and Accessibility

On the security side, the model has updated safeguards to prevent misuse in chemical, biological, radiological, and nuclear fields. These measures, developed against cyberattack risks, are aligned with bioresilience and cybersecurity programs.

Gemini 3.7 Flash is offered at a price of $0.75 for 1 million entry tokens and $3.75 for 1 million exit tokens until the end of 2026. This pricing is half the launch price of the previous model.

The model is now being added to the Spark feature in the Gemini app, requiring an AI Pro or Ultra subscription. It is also accessible via Google Antigravity, AI Studio, Android Studio, Gemini Enterprise Agent Platform, and the Gemini Enterprise app.

Offering more efficient tool usage with Google Workspace applications, the model makes the personal assistant experience more functional for information-driven tasks. What do you think about these performance improvements offered by the new Gemini 3.7 Flash model?