On Tuesday, Google DeepMind unveiled its new models: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. The Gemini 3.6 Flash serves as Google’s primary model, offering enhanced performance in programming, knowledge work, and multimodal tasks while cutting token consumption by as much as 17%, thus making it more economical compared to its predecessor, the 3.5 Flash.
The Gemini 3.5 Flash-Lite is touted as the most economical model in this lineup, while the 3.5 Flash Cyber has been specifically optimized to identify and resolve cybersecurity threats at a reasonable price. This particular model will be available only to governmental entities and trusted partners through a limited-access pilot program.
Google emphasizes that these new releases aim to provide enhanced efficiency, reduced latency, and greater reliability for those developing AI agents on a large scale.
The significance of this launch extends beyond the introduction of cost-effective, faster models tailored for coding, efficiency, and cybersecurity; it also underscores the absence of a widely anticipated update to Google’s flagship model, Gemini Pro, which has not received updates since February.
Since that last update, OpenAI has released GPT-5.5 and initiated the rollout of GPT-5.6, while Anthropic has introduced Claude Opus 4.8 and Claude Sonnet 5, as well as expanded access to its frontier Fable 5 model, showcasing the rapid pace of development among competing labs.
In May, Google hinted at the forthcoming Pro version as part of the 3.5 Flash release, stating that it was already in internal use and would likely roll out the following month. However, recent reports indicated that Google is experiencing internal delays concerning the launch of the 3.5 Pro due to difficulties in meeting performance benchmarks.
The Gemini Pro models are typically regarded as Google’s premium offerings for intricate reasoning and coding tasks, in contrast to the Flash models, which prioritize affordability and quicker response times for production use.
On Tuesday, Logan Kilpatrick, product lead at Google DeepMind, mentioned that Gemini 3.5 Pro is currently undergoing tests with partners and is aiming for a prompt launch. He also revealed that the team has initiated the most ambitious pre-training effort to date for Gemini 4.
