ARK-ASR-3B

Home/Models/Audio8/ARK-ASR-3B

Audio8/ARK-ASR-3B

kind: original
quant: bf16 (16-bit)
family: misc
modality: automatic-speech-recognition
root model: AutoArk-AI/ARK-ASR-3B
provider: Audio8
license: apache-2.0
language: zh, en, de, ja, fr, ko, es, pl, it, ro, hu, cs
context: —
library: transformers
vRAM (GB): —
size on disk: —
downloads: 6,921
likes: 74
created: —
last modified: 2026-06-27T14:45:58+00:00 (2026)
huggingface: https://huggingface.co/Audio8/ARK-ASR-3B
tags: transformerssafetensorsarkasrtext-generationautomatic-speech-recognitionspeechaudiopytorchvllmark-asrcustom_codezhendejafrkoesplitrohucsnl

benchmark	category	score	conf	evidence
WER AMI	speech-asr	8.79	high	`\| ARK-ASR-3B \| 8.79% \| 8.23% \| 6.98% \| 1.03% \| 2.35% \| 2.46% \| 5.47% \| 5.04% \|`
WER Earnings-22	speech-asr	8.23	high	`\| ARK-ASR-3B \| 8.79% \| 8.23% \| 6.98% \| 1.03% \| 2.35% \| 2.46% \| 5.47% \| 5.04% \|`
WER GigaSpeech	speech-asr	6.98	high	`\| ARK-ASR-3B \| 8.79% \| 8.23% \| 6.98% \| 1.03% \| 2.35% \| 2.46% \| 5.47% \| 5.04% \|`
WER VoxPopuli	speech-asr	5.47	high	`\| ARK-ASR-3B \| 8.79% \| 8.23% \| 6.98% \| 1.03% \| 2.35% \| 2.46% \| 5.47% \| 5.04% \|`
WER SPGISpeech	speech-asr	2.46	high	`\| ARK-ASR-3B \| 8.79% \| 8.23% \| 6.98% \| 1.03% \| 2.35% \| 2.46% \| 5.47% \| 5.04% \|`

none scraped.