Google releases three new Gemini models — but no 3.5 Pro

On Tuesday, Google DeepMind released Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. Gemini 3.6 Flash is Google’s “workhorse model” that promises improved capabilities successful coding, cognition work, and multimodal show portion reducing token usage by up to 17%, making it cheaper than its predecessor 3.5 Flash. 

Gemini 3.5 Flash-Lite is the astir cost-effective exemplary successful the class, and 3.5 Flash Cyber is simply a specialized exemplary that was fine-tuned for uncovering and fixing cybersecurity vulnerabilities astatine a decent terms point. This exemplary volition beryllium exclusively disposable to governments and trusted partners arsenic portion of a constricted entree aviator program, according to Google. 

Google says the absorption connected these releases is to present efficiency, latency, and reliability to customers that are gathering AI agents astatine scale. 

The motorboat is notable not conscionable for what Google shipped — cheaper, faster models optimized for coding, efficiency, and cybersecurity — but for what it didn’t. The update doesn’t see the long-anticipated update to Google’s flagship model, Gemini Pro, which was past updated successful February. 

In the clip since that launch, OpenAI has released GPT-5.5 and begun rolling retired GPT-5.6, portion Anthropic has launched Claude Opus 4.8, Claude Sonnet 5, and expanded entree to its frontier Fable 5 model, highlighting the aggravated merchandise gait of the rival labs.

Google teased the merchandise of Pro arsenic portion of the 3.5 Flash merchandise successful May, saying the Pro mentation was “already being utilized internally, and we look guardant to rolling it retired adjacent month.” Last week, Bloomberg reported that Google was facing interior delays successful launching the 3.5 Pro arsenic it struggled to conscionable interior show goals.

Gemini Pro models are mostly Google’s highest-capability offerings for analyzable reasoning and coding tasks, portion Flash models prioritize little outgo and faster effect times for accumulation applications.

Google DeepMind merchandise pb Logan Kilpatrick said Tuesday that the institution is presently investigating Gemini 3.5 Pro with partners and hopes to “land soon.” He besides noted that the squad has started its astir ambitious pre-training tally yet for Gemini 4.