AI IQ Rankings Reveal Tight Race Among Frontier Models

AI IQ Scores Map Frontier Model Landscape

Startup AI IQ has plotted more than 50 leading language models on a human‑style IQ curve, sparking both applause and ire across the tech community. The site averages twelve benchmark results across four reasoning dimensions—abstract, mathematical, programmatic and academic—to produce a single IQ figure, then adds an emotional‑intelligence (EQ) overlay.

Top of the curve compresses rapidly

Mid‑May data places OpenAI’s GPT‑5.5 at ↑ 136, edging out Anthropic’s Opus 4.7 (≈132) and Google’s Gemini 3.1 Pro (≈131). Reuters notes that such clustering mirrors earlier findings from independent Mensa‑style rankings.

“The visual makes a bewildering market almost readable,” writes tech commentator Thibaut Mélen.

Critics warn that collapsing jagged capabilities into one number creates a false sense of precision. A researcher on X argues that GPT‑5.5 still fails ClockBench about half the time, exposing the “jaggedness” problem.

EQ adds a new axis

Anthropic’s Opus 4.7 leads the EQ chart with a score near 132, positioning it in the coveted upper‑right quadrant of the IQ‑vs‑EQ scatter. OpenAI’s models rank high on IQ but trail slightly on EQ, while many Chinese offerings sit in the mid‑range.

Cost versus capability

The IQ‑vs‑effective‑cost plot shows that the most powerful models command premium prices—GPT‑5.5 costs over ↓ $30 per 2 M‑token task—whereas mid‑tier models like DeepSeek‑V3.2 deliver IQs around 115 for under $5.

Enterprise buyers are now forced to orchestrate multi‑model stacks, routing complex queries to high‑IQ, high‑cost services and routine work to cheaper alternatives. The shift echoes strategic re‑thinking that began during the pandemic, when firms first layered AI services to balance cost and performance.

While the methodology remains partially opaque and the IQ metaphor contentious, AI IQ offers the only cross‑provider framework that updates weekly and blends cognitive, emotional and economic dimensions into a single dashboard.

Reported by Julian Reed (Consumer Electronics Expert).