ivrit-ai Whisper large-v3-turbo (Hebrew) — GGML Q8_0

8-bit quantization of ivrit-ai/whisper-large-v3-turbo-ggml for whisper.cpp (Metal), used by the LearnBySpeaking macOS app.

Quantized with whisper.cpp v1.9.5 whisper-quantize ggml-model.bin ggml-ivrit-large-v3-turbo-q8_0.bin q8_0 (874 MB instead of 1.6 GB; same transcriptions on our test sentences).

whisper-server -m ggml-ivrit-large-v3-turbo-q8_0.bin -l he -bs 5 -bo 5

SHA-256: 123a936e686b06d45b52dc1790251a1418841352ca94079ae643d57893ffc9a6

License: Apache-2.0, same as the original model by ivrit.ai. All credit for the model goes to ivrit.ai.