Gemini 3.5 Flash

Gemini 3.5 Flash is the first Gemini 3.5 model to combine frontier-level intelligence with execution capability, and it is especially specialized for performing long-horizon agentic and coding tasks.

Key Performance and Features

• Benchmark Advantage: On key metrics such as Terminal-Bench 2.1 (76.2%) and GDPval-AA (1656 Elo), it outperforms Gemini 3.1 Pro.
• Speed and Cost: Based on output token throughput, it is 4x faster than other frontier models, while costing less than half as much, making it advantageous for complex agentic workflows.
• Multimodal Utilization: Leveraging strong multimodal understanding, it can carry out multi-step tasks such as generating interactive web UIs, implementing games from research papers, and converting legacy code to Next.js.

Availability and Use Cases

• Developers and General Users: It is available to developers through Google AI Studio, Android Studio, and the Gemini API, and is embedded as the default model in the Gemini app and AI Mode in Google Search.
• Personal Agent: It serves as the foundation model for Gemini Spark, a personal AI agent that navigates users' digital lives on their behalf.
• Enterprise Use Cases: Companies such as Shopify, Salesforce, and Databricks are already using this model to build agentic workflows including data analysis, enterprise task automation, and real-time monitoring.

The next model, Gemini 3.5 Pro, is scheduled for release next month.