Specs CalculatorROI CalculatorWorkstation BuilderModel TreemapTools
A compressed version of OpenAI Whisper Large v3 produced via LiteASR low-rank approximation, keeping accuracy close to the original while reducing encoder size. The 'acc' variant optimizes for accuracy over speed.
Access model weights, configuration files, and documentation.
See which devices can run this model and at what quality level.
Lite-Whisper is a family of compressed Whisper checkpoints produced using LiteASR (Kamahori et al., arXiv:2502.20583), a post-training compression technique that replaces dense linear layers in the Whisper encoder with low-rank approximations derived from activation statistics.
View on Hugging Facetransformers-acc checkpoint targets the highest accuracy in the Lite-Whisper family, matching or outperforming Whisper Large v3 while being meaningfully smaller and fasterHigh-quality multilingual ASR (99 languages) in latency- or memory-constrained deployments; a cheaper alternative to running vanilla Whisper Large v3.