See this visualization first on the Voronoi app.
Use This Visualization
Ranked: The Smartest AI Models, by IQ
This was originally posted on our Voronoi app. Download the app for free on iOS or Android and discover incredible data-driven charts from a variety of trusted sources.
Mirror mirror on the wall, who’s the smartest AI of them all?
Artificial intelligence is designed to be intelligent. But how do today’s leading AI models stack up in human IQ terms, and which ones are the smartest?
This infographic ranks the smartest AI models based on their performance on the Mensa Norway IQ test, with data compiled by Tracking AI. The Mensa test is a widely recognized and highly difficult IQ exam used to evaluate human intelligence.
For context, the average human IQ score ranges from 90 to 110, while a score above 130 is typically considered genius-level.
AI with Genius-Level IQ
Topping the chart is OpenAI’s text-only o3 model, scoring a 135 on the Mensa IQ test, which puts it in the “genius” category. As a part of ChatGPT, it’s also among the world’s most popular AI tools.
| Model Name | Mensa Norway IQ Test Score |
|---|---|
| OpenAI o3 | 135 |
| Claude-4 Sonnet | 127 |
| Gemini 2.0 Flash Thinking Exp. | 126 |
| Gemini 2.5 Pro Exp. | 124 |
| OpenAI o4 mini | 122 |
| Claude-4 Opus | 120 |
| Grok-3 Think | 112 |
| DeepSeek R1 | 106 |
| Llama 4 Maverick | 105 |
| OpenAI o1 Pro | 102 |
| DeepSeek V3 | 100 |
| GPT4.5 Preview | 99 |
| Grok-3 | 97 |
| Gemini 2.5 Pro Exp. (Vision) | 96 |
| GPT-4o | 93 |
| OpenAI o4 mini high | 92 |
| Claude-3.7 (Vision) | 91 |
| Bing Copilot | 86 |
| Mistral | 85 |
| OpenAI o1 Pro (Vision) | 83 |
| OpenAI o3 (Vision) | 72 |
| Llama-3.2 (Vision) | 70 |
| GPT-4o (Vision) | 63 |
| Grok-3 Think (Vision) | 60 |
Anthropic’s Claude-4 Sonnet and Google’s Gemini 2.0 Flash Thinking follow closely with IQ scores of 127 and 126, respectively. Furthermore, new iterations like the Gemini 2.5 Pro and OpenAI o4 mini both scored over 120, above the average human IQ range.
Overall, these high scores prove that leading AI models are now operating at high levels of intelligence, with some even surpassing the smartest human minds.
However, what’s more surprising is that all of the top 10 smartest AI models are text-only models that cannot read or process images.
The IQ Contrast Between Text and Vision AI Models
Based on IQ scores, it seems like reasoning through words is still a much stronger suit for AI than interpreting and solving visual images and puzzles.
The bottom five AI models by IQ scores are all multimodal models with the ability to read and process images. In particular, OpenAI’s GPT-4o (Vision) and xAI’s Grok-3 Think (Vision) landed far below the human average, scoring 63 and 60 on the test, respectively.
Still, the results are telling: AI isn’t just mirroring human intelligence, it’s even outscoring us in certain areas of cognition and reasoning.
Learn More on the Voronoi App
See what humans are using AI for in 2025, in this infographic on the Voronoi app.