A smarter model to help you learn, plan, and build like never before. With state-of-the-art reasoning, advanced multimodal understanding, and improved agentic capabilities for the most demanding workflows.
With state-of-the-art reasoning capabilities, Gemini 3.5 Pro handles your most complex tasks with unprecedented depth and nuance.
Smart, concise, direct responses with genuine insight over cliche and flattery. Tackles multi-step problems that require careful logical deduction.
Text, images, video, audio — even code. Gemini 3.5 Pro sets a new state-of-the-art on reasoning with unprecedented depth and nuance across every modality.
Exceptional instruction following with meaningful improved tool use and agentic coding. Ships production-quality code faster than any previous model.
Better tool use. Simultaneous, multi-step tasks. Gemini 3.5 Pro's agentic capabilities can build more helpful and intelligent personal AI assistants.
From real-time data visualization to immersive 3D simulations, Gemini 3.5 Pro turns ideas into interactive reality.
Gemini 3.5 Pro uses advanced reasoning to configure live telemetry streams to build dynamic applications like aerospace dashboards.
Gemini 3.5 Pro codes immersive simulations complete with hand-tracking manipulation and dynamic generative audio.
From terrain generation to traffic flow, Gemini 3.5 Pro uses advanced reasoning to code and assemble the many layers of a simulated city.
Gemini 3.5 Pro understands design intent, converting static SVGs into animated, code-based graphics for faster, cleaner web development.
Gemini 3.5 Pro reasons through the atmospheric tone of a novel to build a modern, personalized portfolio.
Leading results across academic reasoning, agentic coding, agentic search, and multimodal understanding.
| Benchmark | Gemini 3.5 Pro | Gemini 3 Pro | Sonnet 4.6 | Opus 4.6 | GPT-5.2 |
|---|---|---|---|---|---|
| Humanity's Last Exam | 44.4% | 37.5% | 33.2% | 40.0% | 34.5% |
| ARC-AGI-2 | 77.1% | 31.1% | 58.3% | 68.8% | 52.9% |
| GPQA Diamond | 94.3% | 91.9% | 89.9% | 91.3% | 92.4% |
| Terminal-Bench 2.0 | 68.5% | 56.9% | 59.1% | 65.4% | 54.0% |
| SWE-Bench Verified | 80.6% | 76.2% | 79.6% | 80.8% | 80.0% |
| LiveCodeBench Pro (Elo) | 2887 | 2439 | — | — | 2393 |
| BrowseComp Agentic Search | 85.9% | 59.2% | 74.7% | 84.0% | 65.8% |
| MMMU-Pro | 80.5% | 81.0% | 74.5% | 73.9% | 79.5% |
| MMMLU Multilingual | 92.6% | 91.8% | 89.3% | 91.1% | 89.6% |
Higher is better. Source: Google DeepMind evaluation methodology. Benchmarks run with no tools unless otherwise noted.
Our most capable model for agentic workflows, advanced coding, and complex multimodal tasks.
input tokens · 64K output tokens
Text · Image · Video · Audio · PDF
Text
Get started with Gemini 3.5 Pro and bring your most ambitious ideas to life.
Get Started