Google DeepMind's Gemini 3 Sets New Reasoning Benchmark, Surpassing GPT-5.3
Gemini 3 smashes academic exams, outperforming GPT-5.3 Codex on multimodal reasoning.
Deep Dive
Gemini 3 is our most intelligent model yet, with state-of-the-art reasoning to help you learn, build, and plan anything. It features GPT-5.3-Codex Thinking (xhigh) and Humanity's Last Exam Academic reasoning (full set, text + MM).
Key Points
- Achieves top scores on Humanity's Last Exam (full set, text + MM), outperforming GPT-5.3-Codex Thinking (xhigh).
- Delivers state-of-the-art reasoning for learning, building, and planning complex tasks.
- Supports multimodal inputs, enabling analysis of images alongside text for deeper understanding.
Why It Matters
Gemini 3 sets a new standard for AI reasoning, empowering professionals to tackle advanced research and development with confidence.