Parakeet-TDT 0.6B v3

NVIDIA's high-throughput multilingual ASR — 25 European languages with auto language detection, handles 24-minute audio at full attention (3 h with local attention). Built for production batch transcription.

License: CC-BY-4.0 · Context: n/a · Released: August 14, 2025

The decision in five lines

The call: Consider — for voice
Best for: voice
Runs on: 23 hardware picks fit (cheapest: Intel Arc B580 12 GB · $249)
Watch out: Non-European languages (Mandarin, Arabic, Hindi outside its set) — Whisper still wins there.
Evidence: Estimated · last verified April 2026

600M: PARAMETERS
STT: TYPE
—: CONTEXT
~2 GB: VRAM AT Q4

Where we recommend this

Every tier slot in the planner where this model is a top or alternate pick. Pulled live from planner.js — when the planner refreshes, this table stays current.

VOICE · MID

Parakeet-TDT 0.6B v3 (NVIDIA)CC-BY-4.0; 10× faster than Whisper turbo on English + 25 European langs; no CJK/Arabic/Hindi coverage.

The call

NVIDIA's high-throughput multilingual ASR — 25 European languages with auto language detection, handles 24-minute audio at full attention (3 h with local attention). Built for production batch transcription.
When not to use: Non-European languages (Mandarin, Arabic, Hindi outside its set) — Whisper still wins there.

Runner notes

NeMo toolkit natively. ONNX (`istupakov/parakeet-tdt-0.6b-v3-onnx`) and CoreML + Apple Neural Engine variants for edge. Runs on Apple Silicon via CoreML/ANE.

License: CC-BY-4.0
Released: August 14, 2025
Maker: NVIDIA
Model card: huggingface.co/nvidia/parakeet-tdt-0.6b-v3 →

Hardware that fits

Every hardware pick whose memory fits this model at the quant we recommend. Sorted cheapest-first — the top row is your best-value fit. Click through for the full buyer’s guide.

Next step

Find-by-model — see what hardware runs this→