Models

AssemblyAI offers several state-of-the-art speech recognition models, each optimized for different use cases. Choose the model that best fits your needs based on accuracy, latency, cost, and language requirements.

Pre-recorded models

Streaming models

Choosing the right model

Pre-recorded

Universal-3 Pro

Universal-3 Pro is our most advanced transcription model, delivering state-of-the-art accuracy across 6 languages with powerful prompting capabilities. It supports prompting in plain language for tasks like context-specific transcription, verbatim output, audio tagging, and speaker diarization, giving you fine-grained control to guide transcription results. With keyterms prompting supporting up to 1,000 words, built-in code switching, and multichannel support, Universal-3 Pro is ideal for complex audio scenarios requiring the highest accuracy.

Try Universal-3 Pro here

Universal-2

Universal 2 offers the broadest language coverage of any of our models, supporting high-accuracy transcription across 99 languages with low latency. It supports customization through keyterms prompting (up to 200 words) and includes features like multichannel support, automatic language detection, code switching, speaker diarization, and more. Universal 2 is the go-to choice when you need reliable transcription across diverse languages.

Try Universal-2 here

Streaming

Universal-3 Pro Streaming

The most accurate model for voice agents that demand the highest quality. Best-in-class accuracy with advanced prompting capabilities, including both keyterms prompting and native prompting. Supports native multilingual code switching, entity accuracy, and disfluency detection.

Learn more about Universal-3 Pro Streaming

Universal-Streaming Multilingual

Multilingual transcription at the speed and cost of Universal-Streaming. Same fast performance and competitive pricing as our English model, but with expanded language coverage. Features intelligent endpointing and keyterms prompting support for up to 100 words.

Learn more about Universal-Streaming Multilingual

Universal-Streaming English

The fastest model for real-time English transcription. Optimized for speed and cost-effectiveness for English-only applications. Features ~300ms word-by-word immutable transcripts, intelligent endpointing, and keyterms prompting support for up to 100 words.

Learn more about Universal-Streaming English

Whisper Streaming

Open-source Whisper model enhanced with AssemblyAI’s reliable infrastructure and unlimited scale. Supports 99+ languages at an accessible price point with automatic language detection and non-speech tags.

Learn more about Whisper Streaming

Pricing

For detailed pricing information, visit our pricing page.

Pre-recorded

Streaming

For volume discounts, please reach out to sales@assemblyai.com.

Next steps

• Explore Speech Understanding features like summarization, sentiment analysis, and more
• Learn about prompting: Universal-3 Pro prompting guide | Universal-3 Pro Streaming prompting guide