Side by side

See the difference.

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

Comparison of nvidia/canary-qwen-2.5b
1 modelnvidianvidia/canary-qwen-2.5bRemove ×
What’s different2.5B parameters · Encoder: Conformer-based · Decoder: Qwen2
DateOct 17, 2026Discovered
Listed rateSee provider quotes
Provider endpoints0 endpoints →
AMI-Cleaned WER ↗English short-form; normalization unversioned7.91% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
Earnings22-Cleaned-AA-chunked WER ↗English short-form; normalization unversioned6.41% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
Gigaspeech-Cleaned WER ↗English short-form; normalization unversioned7.8% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
LS Clean WER ↗English short-form; normalization unversioned1.23% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
LS Other WER ↗English short-form; normalization unversioned2.63% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
SPGISpeech WER ↗English short-form; normalization unversioned1.7% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
URGENT2024 WER ↗English short-form; normalization unversioned13.02% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
Voice Arena Monsoon WER ↗English short-form; normalization unversioned3.6% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
Voxpopuli-AA-Cleaned WER ↗English short-form; normalization unversioned4.14% WER ↗nvidia/canary-qwen-2.5bLeaderboard entry
Context window—
Max output—
Parameters2.5B
Audio inputSupported
TranscriptionSupported

Add another model

Search all models →