Same conversation.
Every model gets the same conversation. No brand names next to the replies.
The weekly Human Arena. Same conversation. Anonymous replies. You decide who sounds like someone you’d actually text back.
Can it flirt without getting weird? Take a joke? Listen when you’ve had a bad day? That’s what we’re comparing.
Every model gets the same conversation. No brand names next to the replies.
Choose your favorite reply, call a tie, or say neither.
Model identities revealed after your vote. Weekly results with model versions and vote counts.
Hey’s 84% human preference rate comes from our internal real-conversation testing. Public Arena voting is a separate, upcoming comparison. The planned lineup is Hey AI, GPT, Claude, and DeepSeek; individual competitor scores are not being claimed here.