Side by side

See the difference.

Scores, capabilities and costs for the models you’re considering.

Comparison of Edge0/ARK-ASR-3B
1 modelEdge0Edge0/ARK-ASR-3BRemove ×
What’s different3.75B parameters · 19 languages · Encoder: Whisper · Decoder: Qwen2
DateOct 17, 2026Discovered
Listed rateSee provider quotes
Provider endpoints0 endpoints →
AMI-Cleaned WER ↗English short-form; normalization unversioned7.89% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
Earnings22-Cleaned-AA-chunked WER ↗English short-form; normalization unversioned8.07% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
Gigaspeech-Cleaned WER ↗English short-form; normalization unversioned6.94% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
LS Clean WER ↗English short-form; normalization unversioned1.04% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
LS Other WER ↗English short-form; normalization unversioned2.37% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
SPGISpeech WER ↗English short-form; normalization unversioned2.46% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
URGENT2024 WER ↗English short-form; normalization unversioned13.29% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
Voice Arena Monsoon WER ↗English short-form; normalization unversioned3.48% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
Voxpopuli-AA-Cleaned WER ↗English short-form; normalization unversioned3.55% WER ↗Edge0/ARK-ASR-3BLeaderboard entry
Context window—
Max output—
Parameters3.75B
Audio inputSupported
TranscriptionSupported

Add another model

Search all models →