The new AI IQ platform ranks GPT-5.5 at 136 on the human IQ scale.

According to monitoring by Beating, developer Ryan Shea has launched a new platform, AI IQ (aiiq.org), which places leading large models on a human IQ bell curve, providing a single number to intuitively answer: “How smart is this model?” Current data on the website shows the following scores among key tested models: • GPT-5.5: 136 (ranked #1) • Claude Opus 4.7 and Gemini 3.1 Pro: tied at 132 • Grok 4.3: 125 • Kimi K2.6: 122 • DeepSeek V4 Pro and Muse Spark: tied at 117 • Qwen3.6: 108 The platform’s algorithm pulls raw scores from 12 public benchmarks, converts them into implied IQ using a calibrated difficulty curve, and calculates an average across four dimensions: abstract reasoning, mathematical reasoning, programming reasoning, and academic reasoning. Missing dimensions are conservatively imputed, ensuring no model appears smarter simply because it was tested on fewer benchmarks. Since all underlying data comes from existing benchmarks, AI IQ does not conduct new tests; its core value lies in translating scattered performance metrics into an intuitive scale anyone can understand. In addition to the overall IQ ranking, the website offers several cross-sectional views. Users can visually filter for cost-efficiency on the “IQ vs. Cost” chart or track each company’s rate of progress on the “Frontier IQ Timeline.” The platform also incorporates emotional intelligence (EQ) scores from EQ-Bench to measure how well models “communicate effectively.”