Cyberax AI Playbook
cyberax.com
← All models
Model · NVIDIA · Speech-To-Text

NVIDIA Parakeet

NVIDIA's STT family. parakeet-tdt-0.6b-v2 tops the HF Open ASR leaderboard for English; parakeet-tdt-0.6b-v3 adds 25-language multilingual support. Very fast on NVIDIA hardware via NeMo.

Modality
Speech-To-Text
License
CC BY 4.0 (Open weights)
Parameter size
0.6B
Released
May 1, 2025
Last verified
June 8, 2026
Runs locally
Yes

Strengths

  • Top-of-leaderboard English accuracy
  • Fast inference on NVIDIA GPUs via NeMo
  • Permissive CC BY 4.0 license

Weaknesses

  • Peak accuracy is the English v2 model; multilingual needs the separate v3 model
  • Tightest performance is in NVIDIA's NeMo runtime

Try it

WhereTypeNotes
Hugging Face (English) weights CC BY 4.0 — English flagship
Hugging Face (multilingual) weights 25-language multilingual

Official sources

Change log

  • — Repointed from parakeet-tdt-1.1b to current flagship parakeet-tdt-0.6b-v2 (English, 2025-05-01); added parakeet-tdt-0.6b-v3 (multilingual, 2025-08-14).
  • — Initial entry.

Related models

GPT-4o Transcribe

OpenAI · Speech-To-Text

OpenAI's hosted speech-to-text model, built on GPT-4o. The API-recommended transcription model, with lower word-error rate and better language recognition than the original Whisper API.

Proprietary 16k ctx
Try on: OpenAI API
Released Mar 2025 · Verified 2026-06-08

Whisper large-v3 Turbo

OpenAI · Speech-To-Text

OpenAI's distilled Whisper variant. ~8× faster than large-v3 with most of the accuracy retained — the practical default for high-throughput STT pipelines.

MIT Open Runs locally
Try on: Hugging Face · Replicate · Groq
Released Oct 2024 · Verified 2026-06-08

Whisper large-v3

OpenAI · Speech-To-Text

High-accuracy multilingual speech-to-text. Best-in-class for non-English audio; the de-facto open baseline.

MIT Open Runs locally
Try on: Hugging Face · Replicate · Groq
Released Nov 2023 · Verified 2026-06-08

Distil-Whisper

Hugging Face · Speech-To-Text

Hugging Face's distillation of Whisper (distil-large-v3.5). Faster than Whisper-large-v3-Turbo on long-form audio at small accuracy cost; English-only — pick when you don't need multilingual.

MIT Open Runs locally
Try on: Hugging Face
Released Nov 2023 · Verified 2026-06-08

Vosk

Alpha Cephei · Speech-To-Text

An offline, lightweight speech recognition toolkit. Runs on phones, Raspberry Pi, and embedded devices — the right choice when Whisper is too heavy.

Apache 2.0 Open Runs locally
Try on: Vosk models · GitHub
Released Jan 2020 · Verified 2026-06-08