Google launches cheaper Gemini models, delays flagship Pro

Google DeepMind released three new AI models on Tuesday: Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber. The releases target businesses building AI agents at scale, prioritizing efficiency, latency, and reliability over frontier-level capabilities.

For European companies looking to automate workflows without unsustainable compute costs, the new 3.6 Flash serves as Google’s updated workhorse. It offers improved coding and knowledge work performance while cutting token usage by up to 17% compared to the previous 3.5 Flash, directly reducing operating expenses. The 3.5 Flash-Lite takes this further as the most cost-effective option in its class.

The 3.5 Flash Cyber model addresses a specific enterprise vulnerability. Fine-tuned to identify and fix cybersecurity flaws, it will be offered only to governments and trusted partners in a limited pilot program. This restricted access approach is likely to draw immediate interest from European public sector institutions and critical infrastructure operators facing strict regulatory and security mandates.

However, the product launch is defined as much by what is missing as by what arrived. Google did not release an update to its flagship Gemini Pro, which handles complex reasoning tasks and has not been refreshed since February. Its absence is a notable shift in the competitive landscape.

In the months since that February update, rival labs have moved aggressively. OpenAI has released GPT-5.5 and begun rolling out GPT-5.6, while Anthropic launched Claude Opus 4.8, Claude Sonnet 5, and expanded access to its frontier Fable 5 model.

Google originally teased the Pro update in May, stating it was "already being used internally, and we look forward to rolling it out next month." Bloomberg reported last week that the delay stems from Google struggling to meet its own internal performance benchmarks for the model.

Google DeepMind product lead Logan Kilpatrick said Tuesday the company is currently testing Gemini 3.5 Pro with partners and hopes to "land soon." Looking further ahead, Kilpatrick noted the team has begun its most ambitious pre-training run yet for Gemini 4, signaling that the race for the top tier of AI capability is far from settled.