Canary-1B
Multilingual speech recognition and translation under open terms.
About
Canary-1B is NVIDIA's encoder-decoder speech model family for transcription and speech translation, built on a FastConformer encoder with a Transformer decoder. The original release covered English, German, French, and Spanish; the later Canary-1b-v2 expanded to 25 European languages with word-level timestamps and reported strong accuracy at a fraction of the size of competitors. Licensing is the wrinkle: the original weights are CC-BY-NC-4.0 (non-commercial), while the Flash and v2 variants are CC-BY-4.0 and cleared for commercial use. Running it well means buying into NVIDIA's NeMo stack and GPU hardware.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Canary-1b-v2: WER 7.15% (AMI), 10.82% (LibriSpeech Clean); AST COMET 79.3 X→EN, 84.56 EN→X (NVIDIA, 2025).
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (verified tier, metrics-v1). Evidence: NVIDIA NeMo team; v2 covers 25 European languages with timestamps; Original nvidia/canary-1b weights are CC-BY-NC-4.0 (non-commercial); Canary-1B-Flash and Canary-1b-v2 are CC-BY-4.0 (commercial use allowed); WER/COMET figures are NVIDIA-reported; best run inside the NeMo/GPU stack. License: Original weights CC-BY-NC-4.0; Flash and v2 variants CC-BY-4.0. Verify which checkpoint you use..
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.0/10
- Maintenance · recency, releases, issue load
- 7.0/10
- Openness · license + docs
- 6.0/10
- Manual review · sanity read + red-flag check
- 7.0/10
- · NVIDIA NeMo team; v2 covers 25 European languages with timestamps
- · Original nvidia/canary-1b weights are CC-BY-NC-4.0 (non-commercial)
- · Canary-1B-Flash and Canary-1b-v2 are CC-BY-4.0 (commercial use allowed)
- · WER/COMET figures are NVIDIA-reported; best run inside the NeMo/GPU stack