Hume AI Launches Real-World Benchmark for Voice Models

IanAndrewsDC · x · 2026-07-15

Hume AI has introduced Real World VoiceEQ, a new benchmark designed to measure the human-likeness of AI voices.

The benchmark evaluates over 40 open-source and closed-source models across 15+ dimensions and 60+ metrics, relying on more than 1 million genuine human ratings. According to the leaderboard, Google and OpenAI lead the closed-source category, while the well-known ElevenLabs surprisingly ranks lower.

Original post →

More from Models

Models channel →