WorkThe story, in brief
Top Frontier AI Models Top Out At C+ ... Barely Better Than Old Models
Frontier AI models are hitting a ceiling: C+ grades vs. actual experts. The scaling era might be over.
The KeyNews take
Why it matters
New research suggests current frontier models have plateaued in real-world performance relative to human experts, challenging the narrative of continuous capability gains and raising questions about whether scale alone can drive further progress.
The key facts
5 to knowFrontier AI models scoring C+ grades
Performance barely better than older models
Study compares AI performance to actual expert benchmarks
Published May 20, 2026
Source: Forbes (John Koetsier)
Go to the source
Forbes Innovationforbes.com
Publisher excerpt: AI is pretty smart, but not as smart as actual experts, according to a new study ...