Google’s Gemini 3 is its most intelligent AI model yet. The new AI promises to deliver impressive features such as generating interactive 3D visualizations and completing tasks on its own (“agentic” capabilities). But how does it really perform?
The Gemini 3 Pro model boasts upgrades in reasoning and response directness compared to previous models. A key feature is Canvas, a built-in workspace that allows users to ask the AI chatbot to generate code and preview output. The model can also interpret material from various sources at once and handle more complex prompts.
In our testing, Gemini 3 surprised us with its capabilities. We put it through several tests, including generating interactive 3D visualizations. While not perfect, the results were surprisingly good. Gemini 3’s “zero-shot generation” abilities allowed it to complete tasks it hadn’t been trained on.
Overall, Gemini 3 delivers reasonably well – with some caveats. Its features may not meet every expectation, but they are certainly impressive. As we continue to test and refine our understanding of the AI model, we’ll provide updates on its strengths and weaknesses.
Source: https://www.theverge.com/report/825332/google-gemini-3-pro-ai-hands-on-test