Whisper Large-v3 vs Parakeet

The question

Is NVIDIA Parakeet actually faster and better than Whisper Large-v3 for long-form transcription.

Setup

Input was 90 minutes of podcast audio. Two speakers. Light background noise. Both models ran on GPU.

Result

Whisper Large-v3: 5.1% WER, 2:10 runtime. Industry baseline. Parakeet 1.1B: 4.4% WER, 0:42 runtime. Faster and slightly more accurate.

Verdict

Parakeet wins on both axes for English. Whisper still wins for multilingual content.

Migrated from astro_2604b. Needs review.