FermionResearch 发布 164MB 开源语音识别模型 Phonon-2
原标题：FermionResearch/Phonon-2
AI 摘要
FermionResearch 发布开源英语语音识别模型 Phonon-2，基于 NVIDIA parakeet-tdt-0.6b-v3 量化而来，下载体积仅 164 MB，在 Open ASR Leaderboard 七个英语测试集上平均词错误率 5.21%，接近其 2.5 GB 全精度教师模型的 4.96%。该模型采用约 2.1 比特的五级权重量化，可在 Apple Silicon、CPU、H100 及 Android 等设备上运行，并已集成到其 Mac 听写应用 Detta 中。
正文节选
--- license: cc-by-4.0 base_model: nvidia/parakeet-tdt-0.6b-v3 base_model_relation: quantized language: - en library_name: mlx pipeline_tag: automatic-speech-recognition tags: - mlx - apple-silicon - speech-to-text - asr - stt - low-bit - quantization-aware-training - on-device metrics: - wer --- # Phonon-2 Phonon-2 is the most accurate open speech recognition model for English under 900 MB. Across the Open ASR Leaderboard's seven English sets it averages 5.21 % word error, beating models mult