Google DeepMind has unveiled three new additions to its Gemini AI model family: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. The releases, announced on Tuesday, underscore Google's strategic focus on delivering efficiency, reduced latency, and enhanced reliability for organisations developing AI agents at scale. Gemini 3.6 Flash, positioned as Google's new 'workhorse model', promises significant improvements in areas such as coding, general knowledge tasks, and multimodal performance, while simultaneously reducing token usage by up to 17% compared to its predecessor, 3.5 Flash, making it a more cost-effective solution.
Further expanding the Flash series, Gemini 3.5 Flash-Lite has been introduced as the most economical option in its class. A notable inclusion is Gemini 3.5 Flash Cyber, a specialist model meticulously fine-tuned for identifying and rectifying cybersecurity vulnerabilities. This particular model will be accessible exclusively to governments and vetted partners through a restricted access pilot programme, reflecting a targeted approach to critical infrastructure protection and national security applications.
However, the latest announcement is perhaps most significant for what it omitted: the long-anticipated update to Google's flagship Gemini Pro model. The Pro version, last updated in February, is designed for complex reasoning and advanced coding tasks, typically representing Google's highest-capability AI offering. Google had previously hinted at the 3.5 Pro's release in May, suggesting it was already in internal use and would be rolled out 'next month'. Recent reports from Bloomberg indicated that Google has been grappling with internal delays, struggling to meet its performance benchmarks for the model.
The continued absence of Gemini 3.5 Pro contrasts sharply with the aggressive release schedules of Google's key competitors. Since Google's last Pro update, OpenAI has launched GPT-5.5 and commenced the rollout of GPT-5.6, while Anthropic has introduced Claude Opus 4.8, Claude Sonnet 5, and broadened access to its cutting-edge Fable 5 model. This rapid innovation by rivals highlights the intense competition in the frontier AI space and raises questions about Google's ability to maintain pace at the highest end of the market.
Logan Kilpatrick, a product lead at Google DeepMind, acknowledged the ongoing development, stating on Tuesday that Gemini 3.5 Pro is currently undergoing testing with partners and is expected to 'land soon'. He also revealed that the team has embarked on its most ambitious pre-training run to date for the subsequent iteration, Gemini 4, signalling Google's long-term commitment to advancing its AI capabilities despite the current delays.