HF Model & Benchmark Matrix

24,078 variants · 14,984 scores · 120 benchmarks

canary-qwen-2.5b

nvidia/canary-qwen-2.5b

kind
original
quant
bf16 (16-bit)
family
qwen
modality
automatic-speech-recognition
root model
nvidia/canary-1b-flash
provider
nvidia
license
cc-by-4.0
language
en
context
—
library
nemo
vRAM (GB)
—
size on disk
—
downloads
49,755
likes
452
created
—
last modified
2026-04-21T12:31:18+00:00 (2026)
huggingface
https://huggingface.co/nvidia/canary-qwen-2.5b
tags
nemosafetensorsautomatic-speech-recognitionspeechaudioTransformerFastConformerConformerpytorchNeMoQwenhf-asr-leaderboardendataset:Granarydataset:YTCdataset:Yodas2dataset:LibriLightdataset:librispeech_asrdataset:fisher_corpusdataset:Switchboard-1dataset:WSJ-0dataset:WSJ-1dataset:National-Singapore-Corpus-Part-1dataset:National-Singapore-Corpus-Part-6

Benchmark scores (6)

benchmarkcategoryscorecontextconfevidence
WERspeech-asr16.71% WERmedium| % WER | 16.71 | 13.85 | 17.71 | 29.46 |
WER Earnings-22speech-asr10.42low| 2.5.0 | Canary-Qwen-2.5B | 418 | 5.63 | 10.18 | 9.41 | 1.60 | 3.10 | 10.42 | 1.90 | 2.72 | 5.66 |
WER AMIspeech-asr10.18low| 2.5.0 | Canary-Qwen-2.5B | 418 | 5.63 | 10.18 | 9.41 | 1.60 | 3.10 | 10.42 | 1.90 | 2.72 | 5.66 |
WER GigaSpeechspeech-asr9.41low| 2.5.0 | Canary-Qwen-2.5B | 418 | 5.63 | 10.18 | 9.41 | 1.60 | 3.10 | 10.42 | 1.90 | 2.72 | 5.66 |
WER VoxPopulispeech-asr5.66low| 2.5.0 | Canary-Qwen-2.5B | 418 | 5.63 | 10.18 | 9.41 | 1.60 | 3.10 | 10.42 | 1.90 | 2.72 | 5.66 |
WER TED-LIUMspeech-asr2.72low| 2.5.0 | Canary-Qwen-2.5B | 418 | 5.63 | 10.18 | 9.41 | 1.60 | 3.10 | 10.42 | 1.90 | 2.72 | 5.66 |

Discussions (0)

none scraped.

Hugging Face model & benchmark data scraped & matrix-ified. 24,078 variants · 14,984 scores · 120 benchmarks.