Fact · Speech recognition
NVIDIA's model card for Parakeet TDT 0.6B v3 says its transcripts may be less than 100 percent accurate, and that accuracy depends on the language and the characteristics of the input audio.
Checked against its source on October 17, 2026
In the source's own words
“Accuracy varies based on language and characteristics of input audio”
nvidia/parakeet-tdt-0.6b-v3 model card (Hugging Face), NVIDIA, Hugging Face, 2025, Technical Limitations and Mitigation section of the model card.
Open the original- What it measures
- The model card lists the factors as domain, use case, accent, noise, speech type and the context of the speech.
- What it does not show
- This is the vendor's own warning, and it gives no figure for any particular setting. Measure the word error rate on recordings like yours before you rely on a model.