On Tuesday, Google DeepMind launched Gemini 3.6 Flash, 3.5 Flash-Lite, and three.5 Flash Cyber. Gemini 3.6 Flash is Google’s “workhorse mannequin” that guarantees improved capabilities in coding, information work, and multimodal efficiency whereas lowering token utilization by as much as 17%, making it cheaper than its predecessor 3.5 Flash.
Gemini 3.5 Flash-Lite is essentially the most cost-effective mannequin within the class, and three.5 Flash Cyber is a specialised mannequin that was fine-tuned for locating and fixing cybersecurity vulnerabilities at a good worth level. This mannequin might be completely obtainable to governments and trusted companions as a part of a restricted entry pilot program, in keeping with Google.
Google says the concentrate on these releases is to ship effectivity, latency, and reliability to prospects which are constructing AI brokers at scale.
The launch is notable not only for what Google shipped — cheaper, sooner fashions optimized for coding, effectivity, and cybersecurity — however for what it didn’t. The replace doesn’t embody the long-anticipated replace to Google’s flagship mannequin, Gemini Professional, which was final up to date in February.
Within the time since that launch, OpenAI has launched GPT-5.5 and begun rolling out GPT-5.6, whereas Anthropic has launched Claude Opus 4.8, Claude Sonnet 5, and expanded entry to its frontier Fable 5 model, highlighting the extraordinary launch tempo of the rival labs.
Google teased the discharge of Professional as a part of the three.5 Flash launch in Might, saying the Professional model was “already getting used internally, and we look ahead to rolling it out subsequent month.” Final week, Bloomberg reported that Google was dealing with inside delays in launching the three.5 Professional because it struggled to satisfy inside efficiency targets.
Gemini Professional fashions are typically Google’s highest-capability choices for complicated reasoning and coding duties, whereas Flash fashions prioritize decrease price and sooner response occasions for manufacturing purposes.
Google DeepMind product lead Logan Kilpatrick said Tuesday that the corporate is at the moment testing Gemini 3.5 Professional with companions and hopes to “land quickly.” He additionally noted that the group has began its most bold pre-training run but for Gemini 4.
If you buy by hyperlinks in our articles, we may earn a small commission. This doesn’t have an effect on our editorial independence.

