> ## Documentation Index
> Fetch the complete documentation index at: https://usefulai.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Best Real-Time Transcription Models in 2026

> Compare the best real-time transcription models in 2026 for voice agents, live captions, meetings, and multilingual streaming, plus local options.

Updated July 12, 2026

Real-time transcription models turn speech into text as you talk, trading off accuracy, latency, price, and how fast they detect when a speaker finishes. Vendor latency claims rarely mean the same thing, so we ranked 15 streaming models on one independent benchmark.

Our primary score is final-transcript accuracy from the Artificial Analysis streaming speech-to-text benchmark, normalized to a 0-100 scale where higher is better. It covers English-language audio only, so treat it as a strong baseline, not the last word: telephony, accents, and other languages can shift the order.

## Best Real-Time Transcription Models

| # | Model | Best for | Score About score (UsefulAI's 0-100 streaming score is normalized from Artificial Analysis final-transcript word error. Higher means more accurate.) | Price About price (Current USD per hour of streaming audio for the represented route. Volume, plan, region, features, and self-hosting can change the price.) | License About license (Proprietary means no public model weights. Open weight means weights are available, though exact licenses and commercial-use terms vary.) |
| -: | - | - | -: | -: | - |
| 1 | [![](https://usefulai.com/images/icons/48/elevenlabs.io.png)ElevenLabs Scribe v2 Realtime](https://elevenlabs.io/realtime-speech-to-text) | Highest-accuracy multilingual streaming | 100 | \$0.39/hour | Proprietary |
| 2 | [![](https://usefulai.com/images/icons/48/cartesia.ai.png)Cartesia Ink 2](https://docs.cartesia.ai/build-with-cartesia/stt/latest) | Accuracy-first voice agents | 100 | \$0.40/hour | Proprietary |
| 3 | [![](https://usefulai.com/images/icons/48/qwen.ai.png)Qwen3 ASR Flash Realtime](https://www.alibabacloud.com/help/en/model-studio/real-time-speech-recognition-user-guide) | Budget multilingual accuracy | 98 | \$0.17/hour | Proprietary |
| 4 | [![](https://usefulai.com/images/icons/48/x.ai.png)Grok Speech-to-Text Streaming](https://docs.x.ai/developers/models/speech-to-text) | Low-cost streaming with turn detection | 96 | \$0.20/hour | Proprietary |
| 5 | [![](https://usefulai.com/images/icons/48/assemblyai.com.png)AssemblyAI Universal-3.5 Pro Realtime](https://www.assemblyai.com/blog/contextual-awareness-in-universal-3-5-pro-realtime) | Tunable accuracy for voice agents | 93 | \$0.45/hour | Proprietary |
| 6 | [![](https://usefulai.com/images/icons/48/soniox.com.png)Soniox v5 Real-Time](https://soniox.com/docs/stt/models) | Best price-to-performance streaming | 89 | \$0.12/hour | Proprietary |
| 7 | [![](https://usefulai.com/images/icons/48/google.com.png)Google Chirp 3 Streaming](https://docs.cloud.google.com/speech-to-text/docs/models/chirp-3) | Broad multilingual coverage | 86 | \$0.24/hour | Proprietary |
| 8 | [![](https://usefulai.com/images/icons/48/openai.com.png)OpenAI GPT Realtime Whisper](https://developers.openai.com/api/docs/models/gpt-realtime-whisper) | Realtime voice-app transcription | 85 | \$1.02/hour | Proprietary |
| 9 | [![](https://usefulai.com/images/icons/48/inworld.ai.png)Inworld STT 1 Realtime](https://docs.inworld.ai/stt/overview) | Cheapest low-latency streaming | 82 | \$0.10/hour | Proprietary |
| 10 | [![](https://usefulai.com/images/icons/48/mistral.ai.png)Mistral Voxtral Mini Transcribe Realtime](https://docs.mistral.ai/models/model-cards/voxtral-mini-transcribe-realtime-26-02) | Self-hostable streaming accuracy | 80 | \$0.36/hour | Open weight |
| 11 | [![](https://usefulai.com/images/icons/48/microsoft.com.png)Azure AI Speech Real-Time Transcription](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-to-text) | Enterprise multilingual transcription | 80 | \$0.40/hour | Proprietary |
| 12 | [![](https://usefulai.com/images/icons/48/nvidia.com.png)NVIDIA Nemotron 3.5 ASR Streaming 0.6B](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b) | Self-hosted streaming ASR | 79 | n/a | Open weight |
| 13 | [![](https://usefulai.com/images/icons/48/aws.amazon.com.png)Amazon Transcribe Streaming](https://docs.aws.amazon.com/transcribe/latest/dg/streaming.html) | Managed enterprise streaming | 74 | \$0.60/hour | Proprietary |
| 14 | [![](https://usefulai.com/images/icons/48/deepgram.com.png)Deepgram Nova-3](https://developers.deepgram.com/docs/models-languages-overview) | Fast voice-agent default | 64 | \$0.25/hour | Proprietary |
| 15 | [![](https://usefulai.com/images/icons/48/deepgram.com.png)Deepgram Flux General EN](https://developers.deepgram.com/docs/flux/quickstart) | Fastest endpointing for agents | 55 | \$0.34/hour | Proprietary |

Show more
Show less

***

![](https://mintcdn.com/usefulai/gagbTTg8B6ywPGFE/images/icons/144/elevenlabs.io.png?fit=max\&auto=format\&n=gagbTTg8B6ywPGFE\&q=85\&s=9125f7ed780225d05b6dde1d4ed0e78f)

## [ElevenLabs Scribe v2 Realtime](https://elevenlabs.io/realtime-speech-to-text)

ElevenLabs

Highest-accuracy multilingual streaming

[Visit ElevenLabs](https://elevenlabs.io/realtime-speech-to-text)

The most accurate streaming model in the benchmark, with clean live partials and wide language support - the default pick when transcript quality matters most.

Score **100**
Price **\$0.39/hour**
License Proprietary
Time to final **0.141s**

Strengths

<ul>
  <li>Top-tier final accuracy paired with unusually clean, stable partial transcripts, so words hold their place as you speak instead of rewriting themselves.</li>
  <li>Language coverage is broad and detection is automatic, which makes it the safest choice when accuracy across many languages is the priority.</li>
</ul>

Tradeoffs

<ul>
  <li>It is proprietary and API-only, with no self-host route, and sits at the pricier end of the field.</li>
  <li>Diarization is comparatively weak, so for clean multi-speaker separation you may prefer AssemblyAI or a dedicated diarization step.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [ElevenLabs Realtime Speech-to-Text API](https://elevenlabs.io/docs/eleven-api/guides/how-to/speech-to-text/realtime).</li>
</ul>

![](https://mintcdn.com/usefulai/O2NMryn9hetXwNok/images/icons/144/cartesia.ai.png?fit=max\&auto=format\&n=O2NMryn9hetXwNok\&q=85\&s=e9196f0c86dbe2f714ba8560b24a79ee)

## [Cartesia Ink 2](https://docs.cartesia.ai/build-with-cartesia/stt/latest)

Cartesia

Accuracy-first voice agents

[Visit Cartesia](https://docs.cartesia.ai/build-with-cartesia/stt/latest)

Ties for the top accuracy spot and adds genuine semantic endpointing, making it one of the strongest picks built specifically for voice agents.

Score **100**
Price **\$0.40/hour**
License Proprietary
Time to final **0.211s**

Strengths

<ul>
  <li>Final-transcript accuracy is at the very top of the field, and its semantic endpointing judges when you have actually finished a thought rather than just paused.</li>
  <li>That combination makes it one of the most convincing streaming models for building responsive voice agents.</li>
</ul>

Tradeoffs

<ul>
  <li>It is English-only, which rules it out for multilingual products, and it is proprietary and API-only.</li>
  <li>If you need broad language coverage, ElevenLabs Scribe v2 or a multilingual model like Qwen3 or Nemotron will serve you better.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Cartesia Speech-to-Text API](https://docs.cartesia.ai/build-with-cartesia/stt/latest).</li>
</ul>

![](https://mintcdn.com/usefulai/Te6KzZ86-OxPuEC2/images/icons/144/qwen.ai.png?fit=max\&auto=format\&n=Te6KzZ86-OxPuEC2\&q=85\&s=77ec239e207f3d7865895d11af129b39)

## [Qwen3 ASR Flash Realtime](https://www.alibabacloud.com/help/en/model-studio/real-time-speech-recognition-user-guide)

Alibaba

Budget multilingual accuracy

[Visit Alibaba](https://www.alibabacloud.com/help/en/model-studio/real-time-speech-recognition-user-guide)

One of the cheapest ways to get near-top final accuracy, as long as you can live with rough, unstable live partials.

Score **98**
Price **\$0.17/hour**
License Proprietary
Time to final **0.476s**

Strengths

<ul>
  <li>You get final accuracy close to the best models here at one of the lowest prices in the field, plus strong multilingual and dialect coverage.</li>
  <li>For high-volume, cost-sensitive transcription where the finished transcript matters more than the live feed, it is hard to beat on value.</li>
</ul>

Tradeoffs

<ul>
  <li>Its live partials are rough and unstable - fine if you only consume the final transcript, but a poor fit for interfaces where users watch words appear as they talk.</li>
  <li>For steady live captions, Soniox v5, Cartesia Ink 2, or ElevenLabs are better.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Alibaba Cloud Model Studio](https://www.alibabacloud.com/help/en/model-studio/real-time-speech-recognition-user-guide).</li>
</ul>

![](https://mintcdn.com/usefulai/sV7VJe4pqO2Le0pu/images/icons/144/x.ai.png?fit=max\&auto=format\&n=sV7VJe4pqO2Le0pu\&q=85\&s=421f0adc6c1e753bf2ad0cd1661abbc7)

## [Grok Speech-to-Text Streaming](https://docs.x.ai/developers/models/speech-to-text)

SpaceXAI

Low-cost streaming with turn detection

[Visit SpaceX AI](https://docs.x.ai/developers/models/speech-to-text)

A cheap newcomer with accurate final transcripts and built-in turn detection, though its live partials lag well behind the accuracy leaders.

Score **96**
Price **\$0.20/hour**
License Proprietary
Time to final **0.373s**

Strengths

<ul>
  <li>Accurate final transcripts and built-in turn detection at a low price, from a newcomer clearly aiming at the voice-agent market.</li>
  <li>If you want inexpensive streaming and mostly care about the committed transcript, it is a credible option worth testing.</li>
</ul>

Tradeoffs

<ul>
  <li>Like Qwen3, its live partials trail the accuracy leaders, so it is weaker for interfaces that display text as you talk.</li>
  <li>It is also very new, so integrations and tooling are thinner than Deepgram's or AssemblyAI's. Proprietary and API-only.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [xAI API](https://docs.x.ai/developers/models/speech-to-text).</li>
</ul>

![](https://mintcdn.com/usefulai/O2NMryn9hetXwNok/images/icons/144/assemblyai.com.png?fit=max\&auto=format\&n=O2NMryn9hetXwNok\&q=85\&s=0f393f743b8cac7b8f01a94439e0ea86)

## [AssemblyAI Universal-3.5 Pro Realtime](https://www.assemblyai.com/blog/contextual-awareness-in-universal-3-5-pro-realtime)

AssemblyAI

Tunable accuracy for voice agents

[Visit AssemblyAI](https://www.assemblyai.com/blog/contextual-awareness-in-universal-3-5-pro-realtime)

A top-accuracy incumbent with a rare accuracy-versus-latency switch, though diarization is a paid, slower add-on rather than a core strength.

Score **93**
Price **\$0.45/hour**
License Proprietary
Time to final **0.445s**

Strengths

<ul>
  <li>Among the most accurate streaming models, with an explicit switch between maximum accuracy and minimum latency that few rivals offer, so you can tune the same model to the job.</li>
  <li>Stable, immutable transcripts make it dependable for live captioning.</li>
</ul>

Tradeoffs

<ul>
  <li>Speaker diarization is a paid add-on and has historically been slower than the core transcription, so multi-speaker work costs more and lags.</li>
  <li>If diarization is central, test it carefully; if raw value matters more, Soniox v5 undercuts it heavily.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [AssemblyAI Streaming API](https://www.assemblyai.com/docs/streaming/getting-started/transcribe-streaming-audio).</li>
</ul>

![](https://mintcdn.com/usefulai/fTqNrv6I1z_YE97n/images/icons/144/soniox.com.png?fit=max\&auto=format\&n=fTqNrv6I1z_YE97n\&q=85\&s=f4c9424dc4bf92e9a0e9d03f15761111)

## [Soniox v5 Real-Time](https://soniox.com/docs/stt/models)

Soniox

Best price-to-performance streaming

[Visit Soniox](https://soniox.com/docs/stt/models)

The value outlier here - near-top accuracy and among the fastest finals at the lowest price, and still oddly absent from most roundups.

Score **89**
Price **\$0.12/hour**
License Proprietary
Time to final **0.054s**

Strengths

<ul>
  <li>It combines near-top accuracy, among the fastest finals in the field, and the lowest price of any highlighted model, which is a genuinely rare mix.</li>
  <li>There is also a first-party mobile app, so it is one of the few here you can try without writing code.</li>
</ul>

Tradeoffs

<ul>
  <li>The main hesitation is maturity: it is a smaller, newer vendor than the incumbents, which matters for risk-averse enterprise buyers.</li>
  <li>Feature depth like advanced diarization is still catching up to AssemblyAI and the larger clouds. Proprietary and API-first.</li>
</ul>

How to access

<ul>
  <li>**App** — Available in [Soniox Mobile App](https://app.soniox.com/help-center/mobile-app/usage/getting-started).</li>
  <li>**API** — Accessible via [Soniox Realtime Speech-to-Text API](https://soniox.com/docs/stt/rt/real-time-transcription).</li>
</ul>

![](https://mintcdn.com/usefulai/Ez-pJDkPpztLE7Cr/images/icons/144/google.com.png?fit=max\&auto=format\&n=Ez-pJDkPpztLE7Cr\&q=85\&s=11e0c725f841c45443116a184b63beac)

## [Google Chirp 3 Streaming](https://docs.cloud.google.com/speech-to-text/docs/models/chirp-3)

Google

Broad multilingual coverage

[Visit Google](https://docs.cloud.google.com/speech-to-text/docs/models/chirp-3)

Broad language coverage from a major cloud ASR, but slow finalization and steep list pricing make it hard to recommend for latency-sensitive work.

Score **86**
Price **\$0.24/hour**
License Proprietary
Time to final **1.276s**

Strengths

<ul>
  <li>Very broad language coverage plus the enterprise controls - data residency, regional endpoints, compliance - that regulated products often require.</li>
  <li>As a pure recognition model it is accurate across a wide multilingual range, which is its real reason to exist.</li>
</ul>

Tradeoffs

<ul>
  <li>Finalization is the slowest of any model here, which disqualifies it for latency-sensitive voice agents, and its list pricing is steep until you reach very high volume.</li>
  <li>For real-time work, Soniox v5 or Deepgram are far better fits.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Google Cloud Speech-to-Text API](https://docs.cloud.google.com/speech-to-text/docs/streaming-recognize).</li>
</ul>

![](https://mintcdn.com/usefulai/Te6KzZ86-OxPuEC2/images/icons/144/openai.com.png?fit=max\&auto=format\&n=Te6KzZ86-OxPuEC2\&q=85\&s=745b8837f7535bc53cd70fc2f7024d58)

## [OpenAI GPT Realtime Whisper](https://developers.openai.com/api/docs/models/gpt-realtime-whisper)

OpenAI

Realtime voice-app transcription

[Visit OpenAI](https://developers.openai.com/api/docs/models/gpt-realtime-whisper)

Solid, general-purpose realtime transcription that you pay a heavy premium for - most teams should downshift to a cheaper OpenAI transcribe model.

Score **85**
Price **\$1.02/hour**
License Proprietary
Time to final **0.688s**

Strengths

<ul>
  <li>General-purpose accuracy that holds up well across everyday speech, delivered through a mature, well-documented realtime interface built for conversational voice apps.</li>
  <li>For products that want transcription which just works without configuration, it is a dependable default.</li>
</ul>

Tradeoffs

<ul>
  <li>It is by far the most expensive option here, and it does not lead on accuracy or latency to justify that premium.</li>
  <li>Most teams should drop to a cheaper transcribe model such as GPT-4o Transcribe, or move to Soniox v5 for value.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [OpenAI Realtime API](https://developers.openai.com/api/docs/models/gpt-realtime-whisper).</li>
</ul>

![](https://mintcdn.com/usefulai/52KaILwddzz5TNZ_/images/icons/144/inworld.ai.png?fit=max\&auto=format\&n=52KaILwddzz5TNZ_\&q=85\&s=3671789f084bb60b5e580245818c545a)

## [Inworld STT 1 Realtime](https://docs.inworld.ai/stt/overview)

Inworld

Cheapest low-latency streaming

[Visit Inworld](https://docs.inworld.ai/stt/overview)

The cheapest option on this list pairs low latency with built-in turn detection, a strong budget pick for real-time voice work.

Score **82**
Price **\$0.10/hour**
License Proprietary
Time to final **0.082s**

Strengths

<ul>
  <li>The lowest price on the list combined with very low latency is a strong pairing, and built-in turn detection means you get endpointing without bolting on a separate voice-activity step.</li>
  <li>For cost-conscious real-time voice work, it punches above its price.</li>
</ul>

Tradeoffs

<ul>
  <li>Accuracy is mid-pack rather than class-leading, so for demanding transcription you will want ElevenLabs, Cartesia, or AssemblyAI.</li>
  <li>It is proprietary and API-only, and as a newer entrant its ecosystem is thinner than the incumbents'.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Inworld Speech-to-Text API](https://docs.inworld.ai/stt/overview).</li>
</ul>

![](https://mintcdn.com/usefulai/Te6KzZ86-OxPuEC2/images/icons/144/mistral.ai.png?fit=max\&auto=format\&n=Te6KzZ86-OxPuEC2\&q=85\&s=d5217a212a6b3a169c776cdf75f6e60b)

## [Mistral Voxtral Mini Transcribe Realtime](https://docs.mistral.ai/models/model-cards/voxtral-mini-transcribe-realtime-26-02)

Mistral

Self-hostable streaming accuracy

[Visit Mistral](https://docs.mistral.ai/models/model-cards/voxtral-mini-transcribe-realtime-26-02)

The standout open-weight pick - it reaches hosted-grade accuracy at low latency and can run on your own GPU, minus diarization and turn detection.

Score **80**
Price **\$0.36/hour**
License Open weight
Time to final **0.682s**

Strengths

<ul>
  <li>The strongest open-weight option here: it reaches hosted-grade accuracy at low, configurable latency and is genuinely multilingual.</li>
  <li>You can call the hosted API or run the weights yourself, which gives you privacy and data control while shifting cost to your own hardware.</li>
</ul>

Tradeoffs

<ul>
  <li>There is no diarization or turn detection in the realtime model, so voice agents need extra components around it. Self-hosting needs a capable GPU, not a laptop, so "open" does not mean effortless.</li>
  <li>Cartesia or Deepgram Flux handle turn-taking natively.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Mistral Audio Transcription API](https://docs.mistral.ai/studio-api/audio/speech_to_text).</li>
  <li>**Run locally** — If you have a high-end machine, you can run it with [vLLM](https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602) after downloading weights from [Hugging Face](https://huggingface.co/mistralai/Voxtral-Mini-4B-Realtime-2602).</li>
</ul>

![](https://mintcdn.com/usefulai/52KaILwddzz5TNZ_/images/icons/144/microsoft.com.png?fit=max\&auto=format\&n=52KaILwddzz5TNZ_\&q=85\&s=fc9019cc1050a1467e3173bf01ab7caf)

## [Azure AI Speech Real-Time Transcription](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-to-text)

Microsoft

Enterprise multilingual transcription

[Visit Microsoft](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-to-text)

A mature enterprise ASR with deep customization and compliance features, but it trails the newer wave on latency and real-time value.

Score **80**
Price **\$0.40/hour**
License Proprietary
Time to final **0.625s**

Strengths

<ul>
  <li>Deep customization - custom vocabulary, pronunciation, and tuned models - plus broad language support and the compliance and data-handling controls enterprises need.</li>
  <li>For regulated, large-scale deployments that value configurability over raw speed, it remains a serious option.</li>
</ul>

Tradeoffs

<ul>
  <li>It trails the newer wave on latency and, at standard real-time rates, on price, so it is a weak value pick for greenfield projects.</li>
  <li>For faster or cheaper streaming, Soniox v5, Deepgram, or Inworld are stronger. Proprietary and API-based.</li>
</ul>

How to access

<ul>
  <li>**App** — Available in [Microsoft Speech Studio](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-studio-overview).</li>
  <li>**API** — Accessible via [Azure AI Speech API](https://learn.microsoft.com/en-us/azure/ai-services/speech-service/speech-to-text).</li>
</ul>

![](https://mintcdn.com/usefulai/Te6KzZ86-OxPuEC2/images/icons/144/nvidia.com.png?fit=max\&auto=format\&n=Te6KzZ86-OxPuEC2\&q=85\&s=da3a69cf52131d76f5aa096d31f1a718)

## [NVIDIA Nemotron 3.5 ASR Streaming 0.6B](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b)

NVIDIA

Self-hosted streaming ASR

[View on Hugging Face](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b)

The most flexible self-hosted pick - open weights, tunable latency, and multilingual coverage in a compact 0.6B model.

Score **79**
Price **n/a**
License Open weight
Time to final **0.418s**

Strengths

<ul>
  <li>Open weights with runtime-selectable latency let you dial the accuracy-speed trade without swapping models, and it is multilingual and light enough to self-host at high concurrency.</li>
  <li>Running it yourself keeps audio on your infrastructure and makes cost depend on your hardware rather than an API meter.</li>
</ul>

Tradeoffs

<ul>
  <li>You own the deployment: serving, scaling, and updates are on you, which is real work versus a managed API. Peak accuracy trails the top hosted models.</li>
  <li>If you want open weights without the ops, Voxtral Mini offers a hosted API too.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [NVIDIA Build API trial](https://build.nvidia.com/nvidia/nemotron-asr-streaming/api).</li>
  <li>**Run locally** — If you have a high-end machine, you can run it with [NVIDIA NeMo or Speech NIM](https://docs.nvidia.com/nim/speech/latest/asr/deploy-asr-models/nemotron-asr-streaming.html) after downloading weights from [Hugging Face](https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b).</li>
</ul>

![](https://mintcdn.com/usefulai/O2NMryn9hetXwNok/images/icons/144/aws.amazon.com.png?fit=max\&auto=format\&n=O2NMryn9hetXwNok\&q=85\&s=71c98e62e87ab4587dc368301e4ac9e0)

## [Amazon Transcribe Streaming](https://docs.aws.amazon.com/transcribe/latest/dg/streaming.html)

Amazon

Managed enterprise streaming

[Visit AWS](https://docs.aws.amazon.com/transcribe/latest/dg/streaming.html)

A dependable managed streaming service that now trails newer models on accuracy, latency, and price, with little to pull you toward it.

Score **74**
Price **\$0.60/hour**
License Proprietary
Time to final **0.620s**

Strengths

<ul>
  <li>A mature, heavily operated managed service with predictable behavior, custom vocabulary, and the scale and reliability large deployments count on.</li>
  <li>It handles high-volume streaming transcription dependably across a wide set of languages.</li>
</ul>

Tradeoffs

<ul>
  <li>It now trails newer models on accuracy and latency while costing more than most, so there is little reason to start here on the merits.</li>
  <li>For better accuracy, latency, or price, Soniox v5, Deepgram, or ElevenLabs all lead it.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Amazon Transcribe Streaming API](https://docs.aws.amazon.com/transcribe/latest/APIReference/API_streaming_StartStreamTranscription.html).</li>
</ul>

![](https://mintcdn.com/usefulai/Ez-pJDkPpztLE7Cr/images/icons/144/deepgram.com.png?fit=max\&auto=format\&n=Ez-pJDkPpztLE7Cr\&q=85\&s=c38be388669369e23390b67010ac681b)

## [Deepgram Nova-3](https://developers.deepgram.com/docs/models-languages-overview)

Deepgram

Fast voice-agent default

[Visit Deepgram](https://developers.deepgram.com/docs/models-languages-overview)

The long-standing voice-agent default: very fast and reliable on clean audio, though newer models have caught and passed it on accuracy.

Score **64**
Price **\$0.25/hour**
License Proprietary
Time to final **0.066s**

Strengths

<ul>
  <li>Fast, reliable streaming that made it the long-time default for voice agents, with strong tooling, mature SDKs, and consistent low-latency behavior on clean audio.</li>
  <li>For straightforward English voice pipelines, it is still a safe, well-supported workhorse.</li>
</ul>

Tradeoffs

<ul>
  <li>On noisy, accented, or telephony audio its accuracy slips more than the newer leaders, and several models now beat it on final quality.</li>
  <li>If accuracy is the priority, Soniox v5, ElevenLabs, or Cartesia are stronger; for turn-taking, look at Flux.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Deepgram Live Streaming API](https://developers.deepgram.com/docs/live-streaming-audio).</li>
</ul>

![](https://mintcdn.com/usefulai/Ez-pJDkPpztLE7Cr/images/icons/144/deepgram.com.png?fit=max\&auto=format\&n=Ez-pJDkPpztLE7Cr\&q=85\&s=c38be388669369e23390b67010ac681b)

## [Deepgram Flux General EN](https://developers.deepgram.com/docs/flux/quickstart)

Deepgram

Fastest endpointing for agents

[Visit Deepgram](https://developers.deepgram.com/docs/flux/quickstart)

Built for turn-taking rather than raw accuracy - it delivers the fastest finals and fused end-of-turn detection, a deliberate voice-agent trade.

Score **55**
Price **\$0.34/hour**
License Proprietary
Time to final **0.021s**

Strengths

<ul>
  <li>Purpose-built for conversational turn-taking: it fuses transcription with end-of-turn detection and delivers the fastest finals in the field, so agents can respond the moment you actually stop talking.</li>
  <li>For latency-critical voice agents, that focus is the whole point.</li>
</ul>

Tradeoffs

<ul>
  <li>It is English-only in this model, and on final-transcript accuracy it sits at the back of this list, so transcription-quality work is better served elsewhere.</li>
  <li>For higher accuracy, look at Soniox v5 or ElevenLabs; for multilingual, choose a different model entirely.</li>
</ul>

How to access

<ul>
  <li>**API** — Accessible via [Deepgram Flux API](https://developers.deepgram.com/docs/flux/quickstart).</li>
</ul>

***

## How to Choose

When choosing between these models, consider:

* **Access:** Decide first whether you need a managed API, a first-party app, or a self-hosted model, because that choice drives cost, privacy, latency, and setup work. Most models here are API-only. Only Voxtral Mini and Nemotron 3.5 offer a real self-host route, and Soniox and Azure add a first-party app or portal for trying the model without code.
* **Quality:** We use final-transcript accuracy from the Artificial Analysis streaming benchmark, scored 0-100 where higher is better. Watch the partial-versus-final split: a few models (Qwen3, Grok) produce excellent final transcripts but rough live partials, which is invisible in a single accuracy number and matters if users watch text appear as they speak.
* **Price:** We compare on price per hour of streaming audio. Streaming costs more than batch, committed and volume tiers swing prices widely, and features like diarization are often billed on top - so confirm the tier and add-ons before you budget.
* **Time to final transcript:** This is seconds from the end of speech to the final transcript; lower matters most for voice agents, where the practical target is a sub-500ms end-to-end response. Raw latency is only half of it - how well a model detects that a speaker has finished (its endpointing) shapes the felt responsiveness just as much.

For most teams building voice agents, start with Soniox v5 for value, Cartesia Ink 2 or Deepgram Flux when turn-taking is the hard part, and ElevenLabs Scribe v2 when transcript quality outweighs everything. For private or offline deployments, Voxtral Mini and Nemotron 3.5 are the two open-weight picks worth real testing.

***

## Other Models We Considered

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/openai.com.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=21ac965dc6ee5751127e6d435043629b)[OpenAI GPT-4o Transcribe](https://developers.openai.com/api/docs/models/gpt-4o-transcribe) (OpenAI) — Cheaper OpenAI transcription than the realtime model, capable but not latency-first.

![](https://mintcdn.com/usefulai/BapVfRsLI8g6LXdP/images/icons/48/speechmatics.com.png?fit=max\&auto=format\&n=BapVfRsLI8g6LXdP\&q=85\&s=cff366ecfaac9171cfb6b60a74a5fdcd)[Speechmatics Realtime Enhanced](https://docs.speechmatics.com/speech-to-text/models) (Speechmatics) — Strong multilingual and enterprise specialist, but final accuracy trails the leaders.

![](https://mintcdn.com/usefulai/BapVfRsLI8g6LXdP/images/icons/48/smallest.ai.png?fit=max\&auto=format\&n=BapVfRsLI8g6LXdP\&q=85\&s=81d8e83f1c6ea881cdca2824662b495f)[Smallest Pulse](https://docs.smallest.ai/waves/v-4-0-0/model-cards/speech-to-text/pulse) (Smallest.ai) — Very fast finalization, but accuracy and pricing trail the top value picks.

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/openai.com.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=21ac965dc6ee5751127e6d435043629b)[OpenAI Whisper Large v3](https://huggingface.co/openai/whisper-large-v3) (OpenAI) — A great open local baseline, but not truly streaming without wrappers.

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/nvidia.com.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=a4db231e3535440e4b967473be5ffb18)[NVIDIA Parakeet Unified EN 0.6B](https://huggingface.co/nvidia/parakeet-unified-en-0.6b) (NVIDIA) — Local favorite for accuracy and speed, but English-only and GPU-bound.

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/huggingface.co.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=760541dee73285f3eb28c4ced26d9a3c)[Kyutai STT](https://kyutai.org/stt/) (Kyutai) — Open streaming that runs on a typical machine, but no managed API.

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/huggingface.co.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=760541dee73285f3eb28c4ced26d9a3c)[Moonshine v2 Streaming](https://github.com/moonshine-ai/moonshine) (Moonshine AI) — On-device streaming for CPU and phones, but off-benchmark with no managed API.

![](https://mintcdn.com/usefulai/BapVfRsLI8g6LXdP/images/icons/48/gladia.io.png?fit=max\&auto=format\&n=BapVfRsLI8g6LXdP\&q=85\&s=cc9adc87837e6901260b505f97ed2e00)[Gladia Solaria 1 Realtime](https://docs.gladia.io/api-reference/v2/live/init) (Gladia) — A recognizable API option, but the slowest finalization in the benchmark.

![](https://mintcdn.com/usefulai/C5xOaAf4Os-Vu41o/images/icons/48/huggingface.co.png?fit=max\&auto=format\&n=C5xOaAf4Os-Vu41o\&q=85\&s=760541dee73285f3eb28c4ced26d9a3c)[Rev AI Streaming](https://docs.rev.ai/api/streaming) (Rev AI) — An established, affordable baseline, but the weakest accuracy among current models.

***

## Frequently Asked Questions

> [accordion] What is the best real-time transcription model right now?
>
> For pure transcript quality, ElevenLabs Scribe v2 Realtime and Cartesia Ink 2 lead on accuracy. But the model most teams should try first is Soniox v5, which pairs near-top accuracy with among the fastest finals at the lowest price on this list.

> [accordion] What is the best real-time transcription model for most people?
>
> Soniox v5 Real-Time. It is the rare model that is accurate, fast, and cheap at the same time, and there is a first-party app if you want to try it before writing any code. Move to ElevenLabs or Cartesia only when transcript quality has to be the best available.

> [accordion] Is Whisper a real-time transcription model?
>
> Not really. OpenAI's original Whisper is batch-native - it processes fixed audio chunks, so "streaming" wrappers repeatedly recompute overlapping windows, which is slow and jittery. If you want real streaming, use a streaming-native model such as Voxtral Mini Transcribe Realtime, Nemotron 3.5 ASR Streaming, or Kyutai STT.

> [accordion] What is the best real-time transcription model you can run locally?
>
> Voxtral Mini Transcribe Realtime is the strongest open-weight pick, though it needs a capable GPU. Nemotron 3.5 ASR Streaming is smaller but still needs the supported high-end NVIDIA stack. For ordinary-machine or phone deployments, Kyutai STT and Moonshine v2 are the more practical options.

> [accordion] What is the best model for voice agents?
>
> It depends on the hard part. If turn-taking and endpointing are what break your agent, Deepgram Flux fuses transcription with end-of-turn detection and delivers the fastest finals. If you want accuracy plus native endpointing, Cartesia Ink 2 is excellent. For the best overall value, Soniox v5.

> [accordion] Is Deepgram Flux better than Nova-3?
>
> They are built for different jobs. Flux is tuned for conversational turn-taking and the fastest possible finals, which is what voice agents need. Nova-3 is the general-purpose streaming workhorse and scores higher on final-transcript accuracy in our benchmark. Choose Flux for responsiveness, Nova-3 for broader transcription.

> [accordion] Do these benchmark scores match real-world use?
>
> Treat them as a strong starting point, not a guarantee. The benchmark covers English-language audio and does not directly measure 8kHz telephony, multilingual quality, diarization, or full voice-agent latency, so your own workload can reorder the results. Always test the shortlist on your own audio before committing.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.