Parakeet TDT 0.6B v3

NVIDIA's small, fast open-weights speech recognition models; TDT 0.6B v3 covers 25 languages.

Where it ranks

B
Transcription
Models · Local · 4.5% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming)

Vendor: NVIDIA
Family: Parakeet
Released: 14 Aug 2025
Weights: Open: download and run it yourself
License: CC-BY-4.0
Parameters: 0.63B
GPU memory to run locally: ~2 GB (Derived: 0.63B parameters (Hugging Face safetensors count) x 2 bytes = ~1.3 GB of fp16 weights.)
API price: $0.0015 per minute of audio (Artificial Analysis lists $1.5 per 1,000 minutes (Parakeet TDT 0.6B V3, Togetherai, provider Together AI); divided by 1,000.)
Access: Free (open weights)
Last verified: 18 Oct 2026

Price from artificialanalysis.ai, 18 Oct 2026.

Benchmark	Result	Entry on the leaderboard	As of	Source
Artificial Analysis Speech to Text: AA-WER v2 (non-streaming)	4.5%	`Parakeet TDT 0.6B V3, Togetherai` AA-WER v2 (non-streaming), provider Together AI; embedded value 0.0455 converted to percent and rounded; page shows no last-updated date.	18 Oct 2026	artificialanalysis.ai
Coval Benchmarks: Speech-to-Text WER (24h window)	11.2%	`together / parakeet-tdt-0.6b-v3` Coval STT benchmark, mean WER over the trailing 24h window (snapshot generation 2610, as_of 2026-10-19T09:38Z), n=910 clips; rank 30 of 31; read from the public API that feeds https://benchmarks.coval.ai.	19 Oct 2026	benchmarks-api-6wxgp27p2a-ue.a.run.app
Hugging Face Open ASR Leaderboard (English, short-form average WER)	6.1%	`nvidia/parakeet-tdt-0.6b-v3 (fast-gpu-asr)` Average WER over the board's default datasets (English short-form, version 17-10-2026); rank 25; read from the Space's initial table data.	17 Oct 2026	huggingface.co