Skip to content
← Models
Open weightsNVIDIA/10 trustVerified
NeMo Canary logo

NeMo Canary

Multilingual speech recognition and translation family, up to 25 European languages.

Visit official page →

About

NeMo Canary is NVIDIA's family of multilingual speech models. The original Canary was announced February 2024, transcribing English, Spanish, German, and French with punctuation and capitalization plus bidirectional translation between English and the other three. The follow-up Canary-1b-v2 (August 2025) expanded to 25 European languages on the Granary dataset: a 1B-parameter FastConformer encoder plus Transformer decoder achieving strong WER on AMI and LibriSpeech with up to 10x faster inference than Whisper-large-v3 (vendor-reported).

Canary-1b-v2 is available under CC BY 4.0 on Hugging Face (nvidia/canary-1b-v2), making it one of the few permissively licensed multilingual ASR/AST models, with word-level timestamps and punctuation built in. A compact Canary-180M-Flash variant covers the original four languages for lightweight deployment.

Specs

Context window

Pricing

Free quota

API access

Weights

Open weights

Released

February 8, 2024

Benchmarks

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

speech-recognitionspeech-translationtranscription

Trust score — 8/10

/10 trustVerified

Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 84; Trending score 78/100 from researcher assessment.

Sources