Google has unveiled three new lightweight versions of its Gemini AI model, including Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, and a cybersecurity-focused variant, Gemini 3.5 Flash Cyber. However, the company has not provided an updated timeline for the release of its flagship Gemini 3.5 Pro model, which was originally expected in June. The delay is reportedly due to the model falling short of internal performance benchmarks, particularly in coding tasks.
The new models are designed for cost-efficient, high-volume workloads, with Google emphasizing their competitive pricing. Gemini 3.6 Flash, for example, is said to be cheaper per task than rival models from OpenAI, Anthropic, and Chinese competitors. The cybersecurity-focused model, Gemini 3.5 Flash Cyber, is initially available only to governments and trusted partners through a limited-access pilot.
Google's latest releases come amid intense competition in the AI sector. Anthropic recently launched its Mythos 5 and Fable 5 models, which the U.S. government deemed powerful enough to pose a national security risk. OpenAI also released GPT-5.6 earlier this month after a government-requested delay to address security concerns. Chinese rivals, including Moonshot AI and Alibaba, are also gaining momentum, with their models drawing significant demand.
The delay in the Gemini 3.5 Pro model raises questions about Google's ability to keep pace with competitors. CEO Sundar Pichai had previously highlighted the importance of the Pro model as an indicator of Google's AI capabilities. The company's DeepMind AI lab is under scrutiny as rivals continue to push the boundaries of AI performance and applications.