Introducing Gemini 3 Pro: Faster, Safer, and More Capable
Our newest multimodal model delivers sharper reasoning, safer outputs, and better performance for production workloads.
## Why Gemini 3 Pro
Gemini 3 Pro is built for teams that need dependable speed, stronger reasoning, and safer defaults. It handles text, images, and structured inputs with higher quality and lower latency than previous releases.
## What’s new
- Faster responses: Reduced median latency across typical chat and tool-use workloads.
- Better reasoning: Improved instruction-following and multi-step logic for agents and automation.
- Safer by default: Reinforced guardrails and safer refusals on risky content.
- Multimodal upgrades: Stronger text and image understanding with higher fidelity outputs.
## Key capabilities
- Multimodal: Accepts text and images; optimized for code, analysis, and creative tasks.
- Tool use: More reliable function calling and structured output formatting.
- Long context: Handles larger prompts and documents with improved recall.
- Tuned for production: Lower variance in response times and tighter safety alignment.
## Performance highlights
- Latency: Reduced median latency vs. prior generation under standard chat and function-calling workloads.
- Quality: Higher pass rates on reasoning and code benchmarks in internal tests.
- Stability: Better output consistency for structured responses and JSON schemas.
## How to use it
1) Select gemini/3-pro as your model in the dashboard or API.
2) For structured output, include a schema in your prompt or function call definitions.
3) For multimodal, pass images alongside text (base64 or file upload, depending on the endpoint).