AssemblyAI Universal-3.5 Pro Is Now Available on Speechall

August 2, 2026 by Lena Lucille Meyer, Developer Advocate (AI-generated fictional author)

AssemblyAI Universal-3.5 Pro is now available through Speechall for both asynchronous file transcription and live WebSocket transcription at $0.21 per audio hour. Use the model ID assemblyai.universal-3-5-pro with either Speechall endpoint.

With this release, we have deprecated the previous AssemblyAI Universal 3 Pro model, whose Speechall model ID is assemblyai.universal-3-pro. If your integration selects that model explicitly, update the ID to the new version. No removal date is being announced in this post, but we recommend migrating and testing the new model with your own audio.

AssemblyAI Universal-3.5 Pro at a glance

Universal-3.5 Pro is AssemblyAI’s new flagship speech-to-text model for pre-recorded and realtime audio. According to AssemblyAI’s official Universal-3.5 Pro announcement, the release focuses on real-world conversations: multilingual speech, speaker changes, overlapping dialogue, background noise, and terminology that benefits from additional context.

For Speechall users, the essential details are:

• Model name: AssemblyAI Universal-3.5 Pro
• Speechall model ID: assemblyai.universal-3-5-pro
• Price: $0.21 per audio hour
• Availability: asynchronous REST and live WebSocket speech-to-text through Speechall
• Previous model: assemblyai.universal-3-pro, now deprecated

You can browse the available speech-to-text models or use the new identifier directly in an API request.

How to use Universal-3.5 Pro through Speechall

If you already use Speechall, selecting the new AssemblyAI model only requires changing the model ID:

- assemblyai.universal-3-pro
+ assemblyai.universal-3-5-pro

Here is a complete request using the Speechall transcription endpoint:

curl -X POST "https://api.speechall.com/v1/transcribe?model=assemblyai.universal-3-5-pro&output_format=text" 
  -H "Authorization: Bearer $SPEECHALL_API_KEY" 
  -F "[email protected]"

The request uses the same Speechall endpoint and authentication as other providers. This means you can evaluate Universal-3.5 Pro without adding a separate AssemblyAI SDK or redesigning your transcription workflow. See the Speechall getting-started guide for API key setup and your first transcription, or read the speech-to-text API guide for supported upload methods and response formats.

Live transcription over WebSocket

Universal-3.5 Pro is also available through Speechall’s secure WebSocket endpoint for live audio transcription. Select the same model ID when opening the connection:

const wsUrl = new URL('wss://api.speechall.com/v1/transcribe');
wsUrl.searchParams.set('api_key', SPEECHALL_API_KEY);
wsUrl.searchParams.set('model', 'assemblyai.universal-3-5-pro');
wsUrl.searchParams.set('language', 'en');
wsUrl.searchParams.set('output_format', 'json');

const ws = new WebSocket(wsUrl.toString());

ws.onmessage = (event) => {
    const transcription = JSON.parse(event.data);
    console.log(transcription.text);
};

// Send 16 kHz, mono, 16-bit PCM audio chunks after the connection opens.
ws.onopen = () => ws.send(audioChunk);

Browser WebSocket clients pass the Speechall API key as a query parameter because browsers cannot attach a custom Authorization header during the WebSocket handshake. Server-side clients may use the query parameter or a bearer token header. See the Speechall streaming transcription guide for audio requirements, authentication, parameters, and a complete example.

Universal-3.5 Pro versus Universal 3 Pro

The price remains unchanged, so the migration does not increase the per-hour transcription rate. The important integration change is the new explicit model identifier.

Deprecation does not necessarily mean that the old model will stop responding immediately. It means new integrations should choose Universal-3.5 Pro and existing integrations should plan their migration. We are not announcing a removal date for Universal 3 Pro here.

What changed in Universal-3.5 Pro?

AssemblyAI highlights four areas in its release: native code-switching, improved speaker diarization, contextual prompting, and expanded language support. These are AssemblyAI’s descriptions of the underlying model, not independent benchmark results from Speechall.

Native code-switching across 18 languages

Code-switching happens when a speaker moves between languages within the same conversation or even the same sentence. AssemblyAI says Universal-3.5 Pro can transcribe this behavior natively, without requiring a separate pass or manual configuration.

That makes the model relevant to bilingual meetings, international customer calls, interviews, and recordings where speakers naturally mix languages. AssemblyAI reports support for code-switching across 18 languages: English, Spanish, French, German, Italian, Portuguese, Arabic, Danish, Dutch, Finnish, Hebrew, Hindi, Japanese, Mandarin, Norwegian, Swedish, Turkish, and Vietnamese.

Improved speaker diarization

Speaker diarization answers “who said what?” AssemblyAI says Universal-3.5 Pro improves speaker attribution in difficult audio, including short exchanges, interruptions, overlapping speech, and noisy conversations.

This is particularly relevant to meetings, interviews, podcasts, support calls, and other recordings with more than one participant. If diarization is essential to your application, verify the currently exposed model capabilities in Speechall and test with representative recordings before changing production behavior.

Contextual prompting

AssemblyAI also introduced contextual prompting for Universal-3.5 Pro. The idea is to provide domain or conversation context that helps the model recognize specialized terminology, names, products, or topics that may be ambiguous from a short audio segment alone.

This can be useful for technical meetings, industry-specific interviews, healthcare terminology, and product or company names. Optional provider features depend on the parameters currently exposed by the Speechall API, so do not assume every upstream configuration option is available through the unified interface.

Expanded language support

AssemblyAI describes the same 18 languages as having full-accuracy support in Universal-3.5 Pro. For teams working across regions, one multilingual model can simplify evaluation and reduce the need to route each recording to a language-specific model.

Language coverage on paper is only the starting point. Accents, recording conditions, vocabulary, speaker overlap, and your output requirements all affect results. Test the languages and audio conditions that your application actually receives.

What Speechall provides

Speechall provides access to the speech-to-text capabilities of AssemblyAI Universal-3.5 Pro through a single API shared by multiple transcription providers. The model is available for asynchronous file transcription through the REST API and live transcription through the WebSocket API, using the assemblyai.universal-3-5-pro model ID at $0.21 per audio hour.

AssemblyAI’s launch also discusses features such as diarization and contextual prompting. This post does not promise that every optional AssemblyAI feature or parameter is exposed through Speechall, but both asynchronous and realtime WebSocket speech-to-text are supported.

This distinction lets you make an informed choice: use the new model today for speech-to-text, and check the current model catalog for the capabilities exposed through Speechall.

Who should consider upgrading?

Universal-3.5 Pro is worth evaluating if you:

• currently use the deprecated assemblyai.universal-3-pro model;
• transcribe bilingual or multilingual conversations;
• work with meetings, interviews, or calls involving multiple speakers;
• process recordings with interruptions, crosstalk, or background noise; or
• build live captions, voice interfaces, or other realtime transcription experiences; or
• want to compare AssemblyAI’s latest model with other providers through one API.

Even when a newer model is designed to improve broad accuracy, the best migration test uses your own data. Build a small evaluation set that reflects your common languages, accents, microphones, noise levels, file durations, and domain vocabulary. Compare transcripts and any downstream parsing before rolling the new ID out to all traffic.

For a wider evaluation framework, see our guide to best practices for speech API integration.

Migration checklist

1. Replace assemblyai.universal-3-pro with assemblyai.universal-3-5-pro in your model configuration.
2. Keep the existing Speechall transcription endpoint and authentication.
3. Test representative audio, including your hardest languages and recording conditions.
4. Confirm any optional capabilities and output formats your application depends on.
5. Monitor transcript quality and downstream behavior during rollout.

Because the price remains $0.21 per audio hour, there is no price increase associated with selecting the new model on Speechall.

Frequently asked questions

What is the Speechall model ID for AssemblyAI Universal-3.5 Pro?

The model ID is assemblyai.universal-3-5-pro.

How much does AssemblyAI Universal-3.5 Pro cost on Speechall?

It costs $0.21 per audio hour through Speechall.

Is AssemblyAI Universal 3 Pro deprecated?

Yes. Speechall has deprecated assemblyai.universal-3-pro, displayed as AssemblyAI Universal 3 Pro. No removal date is being announced in this post.

How do I migrate to Universal-3.5 Pro?

Change the explicit model ID in your Speechall request or configuration from assemblyai.universal-3-pro to assemblyai.universal-3-5-pro, then test the new model with representative audio before completing your rollout.

Does Universal-3.5 Pro support multiple languages?

AssemblyAI reports full-accuracy support and native code-switching across 18 languages. The supported languages listed in its announcement are English, Spanish, French, German, Italian, Portuguese, Arabic, Danish, Dutch, Finnish, Hebrew, Hindi, Japanese, Mandarin, Norwegian, Swedish, Turkish, and Vietnamese.

Does Speechall expose every Universal-3.5 Pro feature?

Not necessarily. Speechall provides both asynchronous and live WebSocket speech-to-text with the model. Other optional AssemblyAI features depend on the parameters and capabilities currently exposed through the Speechall API.

Can I use Universal-3.5 Pro for realtime transcription through Speechall?

Yes. Use assemblyai.universal-3-5-pro with wss://api.speechall.com/v1/transcribe to stream live audio and receive transcription results over WebSocket. The same model ID is also available for asynchronous file transcription through the Speechall REST API.

Start transcribing with Universal-3.5 Pro

AssemblyAI Universal-3.5 Pro is available on Speechall now for asynchronous and live WebSocket transcription at $0.21 per audio hour. Use assemblyai.universal-3-5-pro for new integrations and migrate away from the deprecated assemblyai.universal-3-pro identifier.

With Speechall, the model joins a unified catalog of speech-to-text providers, so you can try AssemblyAI’s latest model without maintaining another vendor-specific integration. Create an API key, review the available models, and test Universal-3.5 Pro with your own audio.