Google launches three new Gemini models without 3.5 Pro flagship

Google released three new Gemini models and left out a Gemini 3.5 Pro, keeping its AI lineup focused on staged upgrades rather than one headline-grabbing flagship. The omission landed after a fast product cadence: Gemini 2.5 arrived on March 25 as Google’s “most intelligent AI model,” a Gemini 2.5 Pro preview followed on May 6 with better coding performance, and Google updated its thinking-model family again on June 17.

The missing 3.5 Pro matters because Google Cloud documentation describes Gemini 2.5 Pro as Google’s “most advanced reasoning Gemini model,” while Gemini 2.5 Flash-Lite is built for classification, translation, intelligent routing and other cost-sensitive, high-scale operations. That split shows the company is selling Gemini as a family of models with different jobs, not as a single all-purpose answer. For enterprise buyers, the relevant question is less whether Google has the biggest model and more whether it has the right mix of reasoning, speed and cost for production workloads.

Google has spent the past year widening Gemini beyond a standalone chatbot. The company has pushed it through the Gemini app, Vertex AI and Google Cloud, and its Google I/O 2025 updates tied the system to coding, app building and deployment use cases. That matters because enterprise demand often turns on practical deployment details, including latency, throughput, reliability and pricing, not just benchmark scores.

Cost has been part of Google’s strategy before. In a Sept. 24, 2024 developer update, Google said it had reduced Gemini 1.5 Pro pricing and increased rate limits, a sign that affordability and access are part of the competitive fight. The new release pattern suggests Google is still testing where customers will pay for premium reasoning and where they will accept cheaper, specialized models instead.

For Google, the absence of a 3.5 Pro flagship may be as telling as the models it did ship. It leaves room to keep improving the 2.5 line while steering developers and enterprise customers toward the model that fits the task, whether that is advanced reasoning in Gemini 2.5 Pro or high-volume, lower-cost work in Flash-Lite. In a market crowded with OpenAI, Anthropic and other rivals, that is a strategy built around iteration, not one dramatic launch from Mountain View, California.