Google unveiled Gemini 3.5, its newest family of AI models designed to combine frontier-level intelligence with agentic capabilities, enabling developers and enterprises to execute complex, long-horizon workflows more efficiently. The launch begins with Gemini 3.5 Flash, a model focused on coding, multimodal reasoning, and high-speed autonomous task execution.
According to Google, Gemini 3.5 Flash is now available globally through the Gemini app and AI Mode in Search for consumers, while developers can access it through Google Antigravity, Gemini API in Google AI Studio, and Android Studio. Enterprise customers can use the model through the Gemini Enterprise Agent Platform and Gemini Enterprise. Google also confirmed that Gemini 3.5 Pro is currently being tested internally and is expected to launch next month.
Google said Gemini 3.5 Flash delivers performance comparable to large flagship AI models while maintaining the low-latency characteristics of the Flash series. The company highlighted benchmark results, including Terminal-Bench 2.1 at 76.2%, GDPval-AA at 1656 Elo, MCP Atlas at 83.6%, and CharXiv Reasoning at 84.2%. Google added that the model generates output tokens at speeds four times faster than competing frontier models.
The company positioned Gemini 3.5 Flash as a model optimized for large-scale agentic workflows. Google said the system can help automate tasks such as application development, codebase maintenance, and financial document preparation, reducing workflows that previously took days or weeks into significantly shorter timeframes and at lower costs.
Google also emphasized the role of its updated Antigravity harness, which enables Gemini 3.5 Flash to coordinate collaborative subagents capable of executing multi-step workflows and coding operations under supervision. The model also expands multimodal functionality, allowing the creation of richer interactive web interfaces, animations, and graphics.
The company noted that partners are already deploying Gemini 3.5 Flash in production use cases. Google cited examples including banks and fintech firms automating complex workflows and data science teams using the system to uncover insights across large-scale data environments. Shopify was highlighted as a partner using parallel subagents to improve merchant growth forecasting at global scale.
Google additionally introduced Gemini Spark, a personal AI agent powered by Gemini 3.5 Flash. The system is designed to operate continuously, helping users manage digital tasks and take actions under user supervision. Google said Gemini Spark is initially rolling out to trusted testers and will enter beta for Google AI Ultra subscribers in the United States next week.
Gemini 3.5 Flash also powers new Search capabilities, including interactive generative UI experiences and always-on information agents designed to assist users continuously.
Google stated that Gemini 3.5 was developed under its Frontier Safety Framework, with strengthened cyber and CBRN safeguards intended to reduce harmful outputs while minimizing unnecessary refusals for safe queries. The company said it incorporated advanced safety training, mitigations, and interpretability tools to better evaluate and understand model reasoning before responses are generated.
