The best speech-to-text models in 2026
The best speech-to-text models right now are NVIDIA Nemotron 3.5 Transcribe, ElevenLabs Scribe v2 and Voxtral 3 (Transcribe and TTS), out of 21 we have reviewed. Each is ranked by our rating, with its price, our verdict and who it suits.
By ReviewRamble editors · Updated Nov 9, 2026 · How we review
How we ranked: by our rating out of 5, from reviews built on each product's own pricing, documentation and policies plus independent benchmarks where they exist, each checked on the date shown in the review. We take no affiliate commission.
- 1
An open speech model that matches the best open rivals on accuracy and folds diarization, translation and custom vocabulary into one prompt, for $0.12 an hour hosted.
Best for: Teams transcribing recorded calls, meetings or media at volume who want speaker labels and domain vocabulary without stitching several models together.
Price: Open weights under OpenMDW-1.1. DeepInfra: $0.002 per minute of audio ($0.12 an hour); OpenRouter routes to DeepInfra at about the same rate.
Read our NVIDIA Nemotron 3.5 Transcribe review4.5/5 - 2
Scribe v2 transcribes accurately and does the redaction and speaker work for you, but Microsoft now sells better accuracy for less than half the price.
Best for: Meeting, media and clinical products that need speaker labels and redaction as well as an accurate transcript.
Price: API: Scribe v2 and v2 Medical $0.22 per hour of audio; Scribe v2 Realtime $0.39 per hour; keyterm prompting adds $0.05 per hour. Plans from $6/month include hours.
Read our ElevenLabs Scribe v2 review4.4/5 - 3
Near the top of independent transcription tests at a third of a cent a minute, with EU endpoints; the speech side is merely competitive and nothing is open-weight this time.
Best for: Teams transcribing large volumes of meetings, calls or media, especially in Europe, who want top-tier accuracy at low per-minute cost.
Price: Mini Transcribe 3 $0.0035 a minute; Small Transcribe 3 $0.0040 a minute; Mini Transcribe Realtime 3 $0.008 a minute; TTS 3 $13 per million characters (launch prices as reported).
Read our Voxtral 3 (Transcribe and TTS) review4.4/5 - 4
Qwen3-ASR is the open transcriber with the widest reach, and the one to beat for Chinese in any of its forms.
Best for: Developers who need one open model for multilingual and Chinese-dialect transcription, batch or live.
Price: Free under Apache 2.0. Hosted on OpenRouter (DeepInfra, Prime Intellect): 1.7B at $0.0000075 a second of audio (about $0.45 per 1,000 minutes), 0.6B at about $0.20 per 1,000 minutes.
Read our Qwen3-ASR (1.7B and 0.6B) review4.3/5 - 5
Accurate and cheap enough for most transcription, yet no longer the obvious buy on either measure.
Best for: Teams already on the Gemini API who want file and streaming transcription from one vendor at about half a cent a minute.
Price: Transcribe about $0.005 per minute ($2 input / $12 output per 1M tokens); Transcribe Live about $0.009 per minute ($3.50 / $21). Free tier in AI Studio and the API.
Read our Gemini 3.5 Transcribe review4.2/5 - 6
No other open model transcribes 1,600 languages under Apache 2.0, and for the long tail Omnilingual ASR is the default; for English and the big European languages Whisper still holds its ground.
Best for: Teams transcribing languages that commercial speech APIs do not cover, who can run their own GPUs and accept research-grade tooling.
Price: Free under Apache 2.0, commercial use included; the training corpus is CC-BY 4.0. Meta offers no hosted API, so the cost is your own GPU time.
Read our Omnilingual ASR review4.2/5 - 7
MAI-Transcribe-2-Streaming looks like the most accurate live transcription model you can buy, at a mid-market price that only holds until December.
Best for: Teams building live captions, voice agents or call-centre tools on Azure who want top streaming accuracy across many languages.
Price: Introductory $0.54 per hour of audio (about $0.009 per minute) until the end of 2026; Microsoft has not published the price after that.
Read our MAI-Transcribe-2-Streaming review4.1/5 - 8
For transcription on a CPU or a Mac, Parakeet Redux trades a little English accuracy and a lot of noise robustness for a model one-seventh the size and more than twice as fast.
Best for: Developers transcribing reasonably clean speech on CPUs, laptops or Macs who want Parakeet's 25 languages without a GPU.
Price: Free weights under CC BY 4.0 (commercial use with attribution); runs locally through the free moondream Python package.
Read our Moondream Parakeet Redux review4.1/5 - 9
Parakeet v3 is the speed record in transcription, with Whisper-class accuracy and a hard border around Europe.
Best for: Developers transcribing large volumes of European-language audio who prize speed and low cost.
Price: Free under CC BY 4.0. Together AI's hosted version costs $0.0015 a minute ($1.50 per 1,000 minutes).
Read our Parakeet TDT 0.6B v3 review4.1/5 - 10
Phonon-2 is Parakeet at a fifteenth of the size and nearly the same accuracy, which makes good English transcription practical inside an app download.
Best for: App developers who need accurate English speech-to-text on a laptop, phone or CPU server without shipping a multi-gigabyte model.
Price: Weights free under CC-BY-4.0; CLI and code Apache 2.0; Detta app free on Apple Silicon Macs.
Read our Phonon-2 review4.1/5 - 11
Still the open default for telling speakers apart, and a clear step up from 3.1, but its own makers' paid model makes about a third fewer errors on hard audio.
Best for: Engineers building self-hosted meeting, call or podcast transcription who need speaker labels under a permissive licence.
Price: Free under CC BY 4.0. Hosted on pyannoteAI at €0.035 per hour of audio; its paid Precision-3 model is €0.112 per hour (Developer, €19/month) or €0.096 (Starter, €99/month).
Read our pyannote speaker-diarization-community-1 review4.1/5 - 12
Cohere Transcribe is an accurate, permissively licensed English transcriber that leaves timestamps, speaker labels and language detection for you to build.
Best for: Teams transcribing large volumes of single-language, mostly English audio on their own GPUs who want a clean licence.
Price: Free under Apache 2.0 (gated download). Cohere's API is free for testing at 5 requests a minute on trial keys; production limits and Model Vault hourly prices are by quote.
Read our Cohere Transcribe (03-2026) review4.0/5 - 13
Nova-3 is quick and cheap enough for live call audio, but on a neutral test it makes more than twice the errors of the accuracy leaders.
Best for: Contact-centre and live-captioning teams that need fast, high-concurrency transcription of phone audio, including on their own servers.
Price: Pre-recorded $0.0043/min (monolingual) or $0.0052/min (multilingual) pay-as-you-go; streaming $0.0048/min on promotion, $0.0077 regular. Growth plan from $4,000/year is cheaper. $200 free credit.
Read our Deepgram Nova-3 review4.0/5 - 14
Granite Speech 5.0 is the open transcriber to pick when names and jargon matter more than language count.
Best for: Enterprises transcribing English and major European languages with domain vocabulary they can list in advance.
Price: Free under Apache 2.0. A Hugging Face Inference Endpoint is available that scales to zero when idle; its hourly cost depends on the hardware chosen.
Read our Granite Speech 5.0 2B review4.0/5 - 15
A sensible upgrade for OpenAI shops, though the deprecation it rides on takes away speaker labels and timestamps that this model does not give back.
Best for: Teams on the OpenAI API who must leave Whisper or GPT-4o Transcribe before February 2027 and do not need speaker labels or timestamps.
Price: $0.0045 per minute of audio ($4.50 per 1,000 minutes); no Batch support. Streaming gpt-live-transcribe is $0.017 per minute.
Read our GPT Transcribe review3.9/5 - 16
MOSS-Transcribe-Diarize replaces a two-model meeting pipeline with one small one, and its benchmark wins are mostly in Mandarin.
Best for: Teams transcribing long multi-speaker recordings on their own GPUs, especially in Chinese, who want speaker labels without a separate diarisation model.
Price: Free under Apache 2.0. The larger Pro model is offered through MOSI's online playground; API pricing is not published.
Read our MOSS-Transcribe-Diarize 0.9B review3.9/5 - 17
Nemotron 3.5 ASR Streaming is built for running hundreds of live captions on one GPU, and its accuracy is the price of that density.
Best for: Platforms running many simultaneous live transcription or voice-agent sessions on NVIDIA GPUs.
Price: Free under OpenMDW-1.1. Hosted by DeepInfra on OpenRouter at $0.00000333 a second of audio, about $0.20 per 1,000 minutes.
Read our Nemotron 3.5 ASR Streaming 0.6B review3.9/5 - 18
VibeVoice-ASR turns an hour-long meeting into a speaker-labelled transcript in one pass, and every accuracy figure for it is still Microsoft's own.
Best for: Teams transcribing long, multi-speaker recordings on their own hardware who want speaker labels without a separate diarisation step.
Price: Free under the MIT licence. Also available in Azure AI Foundry Labs.
Read our VibeVoice-ASR review3.8/5 - 19
Usable seven-language speech recognition in 16.9 MB, from a vendor that keeps the engine closed and its telemetry on.
Best for: Device makers who need offline voice commands in European languages on hardware too small for Whisper.
Price: Free open weights under Apache 2.0; the prebuilt engine is free to download. Cactus sells fine-tuning and deployment help on its platform; prices are not published.
Read our Whistle review3.8/5 - 20
A dependable default for live captions on OpenAI's stack, priced like a premium model while leaving speaker labels, timestamps and even turn detection for you to build.
Best for: Teams already on OpenAI's Realtime API who need live captions or transcripts in several languages and can do without speaker labels.
Price: $0.017 per minute of streamed audio (about $1.02 an hour), the same as gpt-realtime-whisper. No Batch. The file model gpt-transcribe costs $0.0045 a minute.
Read our GPT Live Transcribe review3.7/5 - 21
Whisper turbo is now the cheap, unrestricted baseline that newer speech models are measured against, not the model that wins.
Best for: Teams that need cheap, permissively licensed transcription they can self-host or buy for cents an hour, and can live with second-tier accuracy.
Price: Free under the MIT licence. Hosted versions are cheap: Artificial Analysis lists Groq at about $0.67 per 1,000 minutes.
Read our Whisper large-v3-turbo review3.7/5
Sources
The main official source behind each ranked review; every review lists its full set.
- NVIDIA Nemotron 3.5 Transcribe: Hugging Face: nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B model card
- ElevenLabs Scribe v2: ElevenLabs API pricing
- Voxtral 3 (Transcribe and TTS): Mistral docs: changelog (Voxtral 3, 20 October 2026)
- Qwen3-ASR (1.7B and 0.6B): Qwen/Qwen3-ASR-1.7B model card
- Gemini 3.5 Transcribe: Gemini API: Gemini 3.5 Transcribe model page
- Omnilingual ASR: facebookresearch/omnilingual-asr on GitHub (README, model table, licence)
- MAI-Transcribe-2-Streaming: Microsoft AI: Our first streaming transcription model debuts at no. 1 on Artificial Analysis
- Moondream Parakeet Redux: Hugging Face: moondream/parakeet-redux model card
- Parakeet TDT 0.6B v3: Parakeet TDT 0.6B v3 model card
- Phonon-2: Hugging Face: FermionResearch/Phonon-2 model card
- pyannote speaker-diarization-community-1: Hugging Face: pyannote/speaker-diarization-community-1 model card
- Cohere Transcribe (03-2026): CohereLabs/cohere-transcribe-03-2026 model card
- Deepgram Nova-3: Deepgram: Introducing Nova-3
- Granite Speech 5.0 2B: ibm-granite/granite-speech-5.0-2b model card
- GPT Transcribe: OpenAI: GPT Transcribe model page
- MOSS-Transcribe-Diarize 0.9B: MOSS-Transcribe-Diarize model card (Hugging Face)
- Nemotron 3.5 ASR Streaming 0.6B: nvidia/nemotron-3.5-asr-streaming-0.6b model card
- VibeVoice-ASR: VibeVoice-ASR model card
- Whistle: Cactus Compute: Whistle, Speech to Text in 16.9 MB
- GPT Live Transcribe: OpenAI: GPT Live Transcribe model page
- Whisper large-v3-turbo: Whisper large-v3-turbo model card
More guides
- Best large language models (LLMs)
- Best AI reasoning models
- Best AI coding models
- Best small and on-device AI models
- Best vision and multimodal AI models
- Best OCR and document AI models
- Best AI image models
- Best AI video models
- Best AI decision models
- Best text-to-speech models
- Best AI music and audio models
- Best embedding models
- Best open-source LLMs
- Best American open-source LLMs
- Best Chinese open-source LLMs
- Best European open-source LLMs