Gemini 3.6 Flash released
Gemini 3.6 Flash is the successor to the version unveiled at Google I/O 2026. According to Google, the model has been improved based on feedback from developers and users.
The biggest change lies in efficiency. According to Google, Gemini 3.6 Flash uses 17 per cent fewer output tokens than Gemini 3.5 Flash and requires fewer reasoning steps and resources to perform complex tasks. This also reduces running costs.
Improvements have also been made in the area of programming. Google says that the model generates more accurate programming code, makes fewer unnecessary changes and produces more reliable code that is ready for use more quickly. Furthermore, the model performs better on knowledge-based and computational tasks. The knowledge base has also been updated from January 2025 to March 2026.
Gemini 3.5 Flash-Lite for fast tasks
Alongside the new Flash model, Google is also introducing Gemini 3.5 Flash-Lite. This variant has been developed for tasks where speed and low latency are important, such as AI-powered searches or document processing.
According to Google, Flash-Lite delivers a clear improvement in quality compared to the earlier 3.1 Flash-Lite version, whilst remaining more cost-effective to use. On several benchmarks, it even outperforms the older Gemini 3 Flash model, including in programming tasks and carrying out practical tasks.
New model for security
Google also announced Gemini 3.5 Flash Cyber. This model has been specifically developed to detect and help resolve security issues in software.
Flash Cyber forms the basis of Google’s CodeMender tool, which deploys multiple AI agents to automatically detect, verify and, where possible, repair vulnerabilities. For the time being, this technology is only available to government bodies and selected partners within a closed-door testing programme.
Availability
Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available from today in the Gemini app. Flash-Lite is also being integrated into Google Search. Developers can get started with both models via Google Antigravity, AI Studio and Android Studio.
Gemini 3.5 Pro and Gemini 4 to follow
Google has also confirmed that Gemini 3.5 Pro is still being tested with partners and will become available as soon as the model is ready for a wider roll-out.
Meanwhile, the company is already looking further ahead. DeepMind is now working on the next generation of AI models and has begun the most ambitious training cycle to date for Gemini 4. Google has not yet announced when that model will be released.
