CALIFORNIA: Google is putting the pedal to the Flash. The technology giant has expanded its Gemini family with three new AI models designed for speed, efficiency and enterprise workloads, while continuing to keep developers waiting for the broader release of its flagship Gemini 3.5 Pro model.
The latest additions to the Gemini Flash lineup focus on handling high-volume tasks with low latency and lower operating costs, rather than competing in the race for increasingly complex reasoning models. The strategy targets developers and businesses building AI agents, automation tools and large-scale enterprise applications.
Leading the announcement is Gemini 3.6 Flash, which succeeds Gemini 3.5 Flash as Google’s primary model for everyday coding, productivity and multimodal tasks. According to Google, the model completes multi-step workflows using around 17 per cent fewer output tokens, reducing both tool calls and internal reasoning steps.
Google has also strengthened the model’s safety mechanisms to better resist jailbreak attempts and reduce misuse in high-risk areas such as cybersecurity and chemical research. Gemini 3.6 Flash is priced at $1.50 per million input tokens and $7.50 per million output tokens.
Alongside it, Google introduced Gemini 3.5 Flash-Lite, a lightweight model built for maximum speed and affordability. Capable of generating up to 350 output tokens per second, Flash-Lite is aimed at latency-sensitive workloads including document ingestion, agentic search and basic computer-use tasks.
Priced at $0.30 per million input tokens and $2.50 per million output tokens, the model is positioned as Google’s lowest-cost offering in the Gemini family. Despite its smaller footprint, the company says it outperforms previous-generation models on benchmarks such as OSWorld-Verified.
Rounding out the launch is Gemini 3.5 Flash Cyber, a specialist model developed exclusively for cybersecurity applications. Integrated with Google’s CodeMender agent, it is designed to analyse large software codebases, validate vulnerabilities in secure sandboxes and generate code patches for human review.
Given the dual-use risks associated with vulnerability detection, Google is limiting access to Flash Cyber through a pilot programme available only to government organisations and selected enterprise partners.
While the Flash lineup expands, the absence of Gemini 3.5 Pro remains notable. Originally expected to launch during the northern hemisphere summer, the flagship model is reportedly still undergoing refinement after internal testing highlighted the need for stronger coding and reasoning performance.
Google said Gemini 3.5 Pro remains in partner testing and will be released more widely once it meets the company’s performance expectations. It also revealed that pre-training work on Gemini 4 has already begun.
The newly launched Gemini 3.6 Flash and Gemini 3.5 Flash-Lite are now available across the Gemini app, Google AI Studio, Android Studio and Google’s Antigravity platform, with Flash-Lite also being integrated into Google Search. Flash Cyber, meanwhile, remains available only through a limited enterprise preview.
