Google drops 3 Gemini models as 3.5 Pro stays in testing

Google on Tuesday released three new AI models built on Gemini 3.5 Flash, pitching them as more token-efficient, faster and more reliable. The update also included a brief status report on Gemini 3.5 Pro, which is still being tested with partners and has not yet been released.

Google’s latest additions are not its flagship models, but each targets a specific use case, including a new version focused entirely on cybersecurity.

Here’s what changed:

• Gemini 3.6 Flash is the company’s new “workhorse” model. Google says it improves coding, knowledge work and multimodal performance, while reducing token usage by up to 17% and lowering the cost per token versus Gemini 3.5.
• Gemini 3.5 Flash-Lite is positioned as Google’s fastest and most cost-effective model, delivering 350 output tokens per second. According to Google, it also significantly outperforms earlier generations on agentic workflows.
• Gemini 3.5 Flash Cyber is a specialist model for finding and fixing vulnerabilities. It works with an infrastructure agent called CodeMender, and Google DeepMind said it is already identifying and patching bugs in Google’s internal Android, Chrome and YouTube codebases.

Both Gemini 3.6 Flash and Gemini 3.5 Flash-Lite support computer use as a built-in tool for more agentic tasks. Google said 3.5 Flash Cyber will initially be limited to governments and trusted partners, with broader access coming later.

As for Gemini 3.5 Pro, Google said only that it is in partner testing and will be released when ready. The company is already looking beyond that: it has begun pretraining for Gemini 4, though no release date was given.

Recommended reading

OpenAI’s New Test Finds Reward-Seeking Grew During RL

Starting Tuesday, Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are available to developers through the Gemini API in Google AI Studio, Android Studio and the Gemini app. 3.6 Flash is also available in Google Antigravity, while 3.5 Flash-Lite is rolling out to Google Search.