Google has introduced the Gemini 3.6 Flash, 3.5 Flash-Lite, and the specialized Gemini 3.5 Flash Cyber models. According to the company, the 3.6 Flash is cheaper than its predecessor, while access to Flash Cyber will initially be limited to a pilot program for governments and trusted partners.

More Efficient and Higher Quality

Google has designated the Gemini 3.6 Flash as the primary working model for coding, data tasks, and multimodal scenarios. The company claims it uses 17% fewer output tokens than the 3.5 Flash, with reductions reaching 65% in specific tests like DeepSWE. The pricing is set at $1.50 per million input tokens and $7.50 per million output tokens.

Google reported improvements in benchmark performance:

  • DeepSWE — 49% compared to 37% for 3.5 Flash;
  • MLE Bench — 63.9% versus 49.7%;
  • OSWorld-Verified — 83.0% against 78.4%.

This has enhanced computer usability, as confirmed by OSWorld testing results. The option is now a built-in tool on the client side via API and the Gemini Enterprise application.

Source: Google.

Focus on Agent Tasks

Google describes the Gemini 3.5 Flash-Lite as the fastest and most affordable model in the lineup. According to Artificial Analysis, it achieves a speed of 350 output tokens per second. The pricing is $0.30 per million input tokens and $2.50 per million output tokens.

Google asserts that this model significantly outperforms the 3.1 Flash-Lite in agent scenarios and surpasses the 3 Flash in several tests, including SWE-Bench Pro (54.2% versus 49.6%) and OSWorld-Verified (74.0% versus 65.1%).

Source: Google.

Cybersecurity in Focus

Additionally, Google announced the Gemini 3.5 Flash Cyber — a specialized model based on the 3.5 Flash designed for identifying and fixing cybersecurity vulnerabilities. In CodeMender, it is utilized as a group of agents that generate a unified report.

According to the company, the model performs competitively on the CyberGym benchmark. Due to its dual purpose, access to Flash Cyber will initially be granted only to governments and trusted partners as part of a limited pilot program.

Source: Google.

The Gemini 3.6 Flash and 3.5 Flash-Lite are already available to developers through the Gemini API in Google AI Studio and Android Studio. For businesses, the models are accessible via the Gemini Enterprise Agent Platform, and the 3.6 Flash has also been integrated into Gemini Enterprise. Regular users can access the models through the Gemini app, with the 3.5 Flash-Lite already being rolled out in Google Search.

Google added that partners are currently testing the flagship Gemini 3.5 Pro, while Gemini 4 is in the pre-training phase.

As a reminder, the company announced the release of the specialized Frozen v2 chip with integrated Gemini architecture.