Live rankings

AI Model Leaderboard

Real-world rankings for LLM, STT and TTS models — computed from arena evaluations on x-n.dev, not vendor benchmarks.

Next recompute in--:--:--
Loading rankings…

Each board ranks by a single metric, aggregated from anonymized arena results across all users and recomputed once a day. Win rate = share of runs matching the expected output; accuracy = word-level match for speech-to-text; latency is the average response time; favorites are community picks. Models need a minimum number of runs before they’re included in the calculation.