AIFast Company Techabout 2 hours ago
Cohere Releases CulturalRiddles Benchmark
4 min read
Cohere Labs researchers tested 28 AI models using 68,154 unique answers from 166 volunteers across 45 countries. Google's Gemini 3.1 Pro Preview earned the highest score at 79.4%, while Alibaba's Qwen 3.5 0.8B finished last with 0.4% accuracy. The study reveals models often guess confidently instead of admitting ignorance, with only 0.7% of the time saying they did not know.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyCohere
