Next-generation multimodal model released February 2025, built for the agentic era with native tool use and enhanced capabilities. Delivers significant performance improvements over 1.5 series with superior speed and 1M token context window.
- Next-generation model built specifically for agentic era applications and autonomous workflows
- Native tool integration eliminating need for external frameworks with superior agentic capabilities
- Enhanced multimodal I/O supporting text, audio, images, video with expanding output modalities
- Live API enabling real-time bidirectional voice and video interactions for dynamic applications
- Major performance improvements across all categories versus Gemini 1.5 with maintained speed
Web Search
Search the web for current info.
Image Generator
Generate images from text prompts.
Model Information
Supported Formats
Gemini 3 ProGoogle's most advanced reasoning model released 2025, featuring configurable thinking levels (low/high) for optimal cost-performance balance. Excels at complex tasks requiring broad knowledge and advanced reasoning with 1M token context window.
Gemini 3 FlashGoogle's fastest frontier AI model built for speed at a fraction of the cost. Outperforms Gemini 2.5 Pro while being 3x faster. Excels at agentic coding (78% SWE-bench) and PhD-level reasoning (90.4% GPQA Diamond).
Gemini 2.5 ProMost advanced thinking model released March 2025, topping LMArena leaderboard with state-of-the-art reasoning capabilities. Features enhanced coding performance with 86.7% on AIME 2025 and 84% on GPQA Diamond, plus configurable thinking budgets up to 32K tokens.