Model reference · open weights

speech-large-114K

speech-large-114K is an open-weight embedding model from PantagrueLLM. speech-large-114K (FP32) weighs 627 MB; the smallest configuration that runs it is RTX 3060 12 GB.

• speech-large-114K is a 313M parameter self-supervised encoder developed by PantagrueLLM for French speech feature extraction.
• The model was pre-trained on diverse French audio data using a data2vec 2.0 teacher-student setup to learn contextualized representations.
• It supports 16 kHz mono audio input and is distributed under an other licence.

Summary of the PantagrueLLM/speech-large-114K model card, 2026-10-01

Hugging Face ↗Paper ↗