Google Launches Gemini 3.5 Flash to Boost AI Performance

Google has launched Gemini 3.5 Flash, a new AI model that is now the default for the Gemini app and AI Mode worldwide [1]. The model is available at no cost to users, marking the first release in the company’s 3.5 model family [1].

Google claims the new model outperforms its predecessor, Gemini 3.1 Pro, on most benchmarks while delivering output speeds four times faster than other frontier models [1]. According to Koray Kavukcuoglu, Google DeepMind’s CTO and Chief AI Architect, the model is specifically optimized for deploying multiple agents simultaneously and managing complex, long-running coding tasks [1]. In internal testing, the company successfully used the model to build a functional operating system from scratch [1].

The release comes as Google attempts to maintain its momentum in a competitive AI market. The company reports that the Gemini chatbot now serves 900 million monthly active users, more than double the 400 million users it recorded at this time last year [1]. To support this growth, Google has significantly ramped up its infrastructure spending, with CEO Sundar Pichai projecting capital expenditures between $180 billion and $190 billion for 2026, a sharp increase from the $31 billion spent in 2022 [1].

While the model is now generally available across Google’s platforms—including the Gemini app, Google AI Studio, and the Gemini Enterprise Agent Platform—the company continues to refine its broader model lineup [1]. A more advanced frontier model, Gemini 3.5 Pro, is currently in internal testing and is scheduled for a public rollout in June [1].

For business users, the shift toward more capable free models like Gemini 3.5 Flash highlights a persistent trade-off in the current AI market. While these tools provide significant productivity gains for tasks like coding and summarization, they often come with usage caps and data governance requirements that vary by account type and organization [2]. As Google integrates these models deeper into its ecosystem, the primary challenge for users remains balancing the speed of these new AI agents against the need for manual verification of their outputs [2].