Model Options

model string Default: base-general

Pre-recorded Streaming:NovaStreaming:Flux

Deepgram’s Model feature allows you to supply a model to use when processing submitted audio. To learn more about the pricing for our different models, see Deepgram Pricing & Plans.

Below are a list of all model and model options that can be used with the Deepgram API.

Examples

https://api.deepgram.com/v2/listen?model=flux-general-en

https://api.deepgram.com/v2/listen?model=flux-general-multi&language_hint=es

Flux is the first conversational speech recognition model built specifically for voice agents. Unlike traditional STT that just transcribes words, Flux understands conversational flow and automatically handles turn-taking. Flux tackles the most critical challenges for voice agents today: knowing when to listen, when to think, and when to speak. The model features first-of-its-kind model-integrated end-of-turn detection, configurable turn-taking dynamics, and ultra-low latency optimized for voice agent pipelines, all with Nova-3 level accuracy.

Flux Multilingual (flux-general-multi) extends Flux to 10 languages with an optional language_hint parameter that biases output toward specified languages. See Language Prompting for details.

Examples

https://api.deepgram.com/v1/listen?model=nova-3

Nova-3 represents a significant leap forward in speech AI technology, featuring substantial improvements in accuracy and real-world application capabilities. The model delivers industry-leading performance with a 53.4% reduction in word error rate (WER) for streaming and 47.4% for batch processing compared to competitors. Nova-3 introduces groundbreaking features including real-time multilingual conversation transcription, enhanced comprehension of domain-specific terminology, and optional personal information redaction. Notably, it’s the first voice AI model to offer self-serve customization, enabling instant vocabulary adaptation without model retraining. In multilingual testing, Nova-3 demonstrated superior performance across all seven tested languages, with particularly strong results showing up to 8:1 preference ratios in certain languages.

• general: Optimized for everyday audio processing.
• medical: Optimized for audio with medical oriented vocabulary.
• pharma: Optimized for audio with pharmaceutical vocabulary, such as drug names, dosages, and medication terminology.

Examples

https://api.deepgram.com/v1/listen?model=nova-2

https://api.deepgram.com/v1/listen?model=nova-2-phonecall

Nova-2 expands on Nova-1’s advancements with speech-specific optimizations to the underlying Transformer architecture, advanced data curation techniques, and a multi-stage training methodology. These changes yield reduced word error rate (WER) and enhancements to entity recognition (i.e. proper nouns, alphanumerics, etc.), punctuation, and capitalization.

• general: Optimized for everyday audio processing.
• meeting: Optimized for conference room settings, which include multiple speakers with a single microphone.
• phonecall: Optimized for low-bandwidth audio phone calls.
• voicemail: Optimized for low-bandwidth audio clips with a single speaker. Derived from the phonecall model.
• finance: Optimized for multiple speakers with varying audio quality, such as might be found on a typical earnings call. Vocabulary is heavily finance oriented.
• conversationalai: Optimized for use cases in which a human is talking to an automated bot, such as IVR, a voice assistant, or an automated kiosk.
• video: Optimized for audio sourced from videos.
• medical: Optimized for audio with medical oriented vocabulary.
• drivethru: Optimized for audio sources from drivethrus.
• automotive: Optimized for audio with automative oriented vocabulary.
• atc: Optimized for audio from air traffic control.

Examples

https://api.deepgram.com/v1/listen?model=nova

https://api.deepgram.com/v1/listen?model=nova-phonecall

Nova is the predecessor to Nova-2. Training on this model spans over 100 domains and 47 billion tokens, making it the deepest-trained automatic speech recognition (ASR) model to date. Nova doesn’t just excel in one specific domain — it is ideal for a wide array of voice applications that require high accuracy in diverse contexts.

• general: Optimized for everyday audio processing. Likely to be more accurate than any region-specific Base model for the language for which it is enabled. If you aren’t sure which model to select, start here.
• phonecall: Optimized for low-bandwidth audio phone calls.

Examples

https://api.deepgram.com/v1/listen?model=enhanced

https://api.deepgram.com/v1/listen?model=enhanced-phonecall

Enhanced models are still some of our most powerful speech-to-text models; they generally have higher accuracy and better word recognition than our base models, and they handle uncommon words significantly better.

• general: Optimized for everyday audio processing. Likely to be more accurate than any region-specific Base model for the language for which it is enabled. If you aren’t sure which model to select, start here.
• meeting beta: Optimized for conference room settings, which include multiple speakers with a single microphone.
• phonecall: Optimized for low-bandwidth audio phone calls.
• finance beta: Optimized for multiple speakers with varying audio quality, such as might be found on a typical earnings call. Vocabulary is heavily finance oriented.

The Enhanced models can be called with the following syntax:

Examples

https://api.deepgram.com/v1/listen?model=base

https://api.deepgram.com/v1/listen?model=base-phonecall

Base models are built on our signature end-to-end deep learning speech-to-text model architecture. They offer a solid combination of accuracy and cost effectiveness in some cases.

• general: (Default) Optimized for everyday audio processing.
• meeting: Optimized for conference room settings, which include multiple speakers with a single microphone.
• phonecall: Optimized for low-bandwidth audio phone calls.
• voicemail: Optimized for low-bandwidth audio clips with a single speaker. Derived from the phonecall model.
• finance: Optimized for multiple speakers with varying audio quality, such as might be found on a typical earnings call. Vocabulary is heavily finance oriented.
• conversationalai: Optimized for use cases in which a human is talking to an automated bot, such as IVR, a voice assistant, or an automated kiosk.
• video: Optimized for audio sourced from videos.

You may also use a custom, trained model associated with your account by including its custom_id.

Examples

https://api.deepgram.com/v1/listen?model=whisper

https://api.deepgram.com/v1/listen?model=whisper-SIZE

Deepgram’s Whisper Cloud is a fully managed API that gives you access to Deepgram’s version of OpenAI’s Whisper model. Read our guide Deepgram Whisper Cloud for a deeper dive into this offering.

Deepgram’s Whisper models have the following size options:

• tiny: Contains 39 M parameters. The smallest model available.
• base: Contains 74 M parameters.
• small: Contains 244 M parameters.
• medium: Contains 769 M parameters. The default model if you don’t specify a size.
• large: Contains 1550 M parameters. The largest model available. Defaults to OpenAI’s Whisper large-v2.

To transcribe audio from a file on your computer using a particular model, run the following curl command in a terminal or your favorite API client.