NVIDIA

Parakeet TDT 0.6B v3

About

Parakeet TDT 0.6B v3 is a 0.6B open-weight model from NVIDIA. NeMo Parakeet speech-to-text transcription. It accepts text, images, and video. Released August 4, 2025.

Parameters
0.6B
Published parameter count.
License
CC-BY-4.0
License on the weights.

Pricing

Token rates are US dollars per 1M tokens. Image rates are per 1K images. Audio rates are per 1M audio seconds.

Rate	Price
Per 1M audio seconds	$16.67

Compare

Parakeet TDT 0.6B v3 is the only transcription model in the catalog. The other rows take similar inputs and do a different job.

Model	What it does	Price	Parameters	Context
Parakeet TDT 0.6B v3	NeMo Parakeet speech-to-text transcription.	$16.67 / 1M audio seconds	0.6B	—
PP-OCRv6	PaddleOCR PP-OCRv6 medium text detection and recognition; scene OCR JSONL on image chat; text returns plain text; document_url fan-out via text.	$0.02 / 1M input	20M	—
EmbeddingGemma 2	Multimodal embeddings: text, images, video, audio, and PDF pages in one 768-d space.	$0.01 / 1M input	740M	8K
Gemini 3.5 Flash Lite	Fastest and cheapest Gemini tier. Multimodal chat; emits no reasoning tokens.	$0.30 / 1M input	—	1M

Benchmarks

Published scores for Parakeet TDT 0.6B v3, from the Parakeet paper.

Evals

A lower score is better.

Reading	Example
Qualitative	Clear structure, grounded in the input
Quantitative	Open ASR avg WER: 6.32.
Cost and performance	Lower listed rate, mid-pack latency

Methods

Methods this model serves. Payload shapes are in the docs.

Method	Returns
transcribe	A transcript

Supported inputs

Audio upload (file)Yes
One file per request, as multipart form data.
Audio URL (url)Yes
A hosted file, in place of file.
Video containerYes
An mp4, mov, webm, or mkv upload or URL is accepted and its audio track is transcribed.
text, image_url, document_urlNo
This model is not a chat model. A chat completion naming it is a 400.

Served over POST /v1/openai/audio/transcriptions. Only method: transcribe. language is an optional ISO-639-1 hint.
Billed per audio second on json and verbose_json replies. See Pricing.
Weights: nvidia/parakeet-tdt-0.6b-v3.

Estimate cost

Hours of audio

$60.01

Quick start

Call this model on the OpenAI-compatible gateway. The model id is already filled in.

from openai import OpenAI

client = OpenAI(
    base_url="https://gateway.vlm.run/v1/openai",
    api_key="<VLMRUN_API_KEY>",
)

with open("sample.mp3", "rb") as audio_file:
    response = client.audio.transcriptions.create(
        model="nvidia/parakeet-tdt-0.6b-v3",
        file=audio_file,
    )

print(response.text)

Response

response_format defaults to json, which carries the transcript and a usage block:

{
  "id": "transcription-3d1df87d0037",
  "object": "audio.transcription",
  "model": "nvidia/parakeet-tdt-0.6b-v3",
  "backend": "nemo",
  "task": "transcribe",
  "duration": 252.726,
  "text": "McKeys Rocks is a rough and tumble area just outside Pittsburgh…",
  "usage": {
    "type": "duration",
    "input_audio_seconds": 252.726,
    "input_tokens": 0,
    "output_tokens": 581,
    "total_tokens": 581,
    "cost": 0.004213
  }
}

usage.cost is the metered charge in USD, and input_audio_seconds is the quantity it was computed from: the clip's duration at $0.06 per audio hour. See Pricing.

Note

Only json and verbose_json carry a usage block. text, srt, and vtt return the transcript alone, so read the cost from one of the JSON formats if you need it.