higgs-audio-v3-stt-v2

Home/Models/bosonai/higgs-audio-v3-stt-v2

bosonai/higgs-audio-v3-stt-v2

kind: original
quant: bf16 (16-bit)
family: misc
modality: automatic-speech-recognition
root model: bosonai/higgs-audio-v3-stt
provider: bosonai
license: apache-2.0
language: —
context: —
library: transformers
vRAM (GB): —
size on disk: —
downloads: 931
likes: 12
created: —
last modified: 2026-04-20T17:39:15+00:00 (2026)
huggingface: https://huggingface.co/bosonai/higgs-audio-v3-stt-v2
tags: transformerssafetensorshiggs_audio_3feature-extractionaudioautomatic-speech-recognitionasrhiggs-audioopen-asr-leaderboardcustom_codeenbase_model:bosonai/higgs-audio-v3-sttbase_model:finetune:bosonai/higgs-audio-v3-sttlicense:apache-2.0region:us

benchmark	category	score	context	conf	evidence
WER	speech-asr	10.03		low	`\| AMI \| 10.03 \|`
WER AMI	speech-asr	10.03	AMI	medium	`\| AMI \| 10.03 \|`
WER Earnings-22	speech-asr	8.95	Earnings-22	medium	`\| Earnings-22 \| 8.95 \|`
WER GigaSpeech	speech-asr	8.16	GigaSpeech	medium	`\| GigaSpeech \| 8.16 \|`
WER VoxPopuli	speech-asr	6.07	VoxPopuli	medium	`\| VoxPopuli \| 6.07 \|`
WER SPGISpeech	speech-asr	3.76	SPGISpeech	medium	`\| SPGISpeech \| 3.76 \|`
WER LibriSpeech	speech-asr	2.8	LibriSpeech other	medium	`\| LibriSpeech other \| 2.80 \|`
WER TED-LIUM	speech-asr	2.76	TED-LIUM	medium	`\| TED-LIUM \| 2.76 \|`
WER LibriSpeech	speech-asr	1.39	LibriSpeech clean	medium	`\| LibriSpeech clean \| 1.39 \|`

none scraped.