Language guide · Mandarin Chinese zh

Best local speech-to-text models for Mandarin Chinese

16 of the 26 reviewed local speech-to-text models explicitly list Mandarin Chinese. For long recordings on a mid-range NVIDIA GPU the matcher's first pick is Whisper turbo; for live transcription it is Nemotron 3.5 ASR Streaming 0.6B; for meetings with speaker labels it is VibeVoice-ASR 7B. 4 models offer native streaming, 3 native speaker diarization and 15 have a permissive license.

Top matches by setup

Computed with the same deterministic matcher as the homepage, restricted to models whose official source lists Mandarin Chinese. To keep the shortlist useful, each row shows the best-ranked model from three different families (for example, only one Whisper size). Scores are decision guidance, not accuracy claims.

Best-ranked model from each of the top three model families, Mandarin Chinese
Setup	#1	#2	#3
Long recordings, 8–12 GB NVIDIA GPU, balanced	Whisper turbo 79/100 · Strong fit	Nemotron 3.5 ASR Streaming 0.6B 75/100 · Strong fit	Qwen3-ASR 0.6B 75/100 · Strong fit
Live transcription (native streaming, speed first)	Nemotron 3.5 ASR Streaming 0.6B 91/100 · Excellent fit	Voxtral Mini 4B Realtime 2602 86/100 · Excellent fit	VibeVoice-ASR Streaming 1.5B 83/100 · Strong fit
Meetings with speaker labels (diarization + timestamps)	VibeVoice-ASR 7B 86/100 · Excellent fit	Whisper medium 84/100 · Strong fit	Qwen3-ASR 0.6B 77/100 · Strong fit
Subtitles with timing, accuracy first	Whisper medium 87/100 · Excellent fit	Qwen3-ASR 0.6B 80/100 · Strong fit	VibeVoice-ASR 7B 79/100 · Strong fit
CPU-only machine, low memory use	VibeVoice-ASR-BitNet 78/100 · Strong fit	Whisper tiny 74/100 · Strong fit	Omnilingual ASR CTC 300M 62/100 · Possible fit

Re-rank for your setup

Change hardware, workload and requirements to re-rank all 16 Mandarin Chinese-capable models in your browser.

Choose by requirement

Live captions and voice input: Native streaming: Nemotron 3.5 ASR Streaming 0.6B, VibeVoice-ASR Streaming 1.5B, Voxtral Mini 4B Realtime 2602 and VibeVoice-ASR-Streaming-7B. Everything else needs a chunking runtime, which adds latency.
Speaker-labelled meetings: Native diarization: VibeVoice-ASR 7B, VibeVoice-ASR Streaming 1.5B and VibeVoice-ASR-Streaming-7B. Other models need a separate diarization pipeline.
Commercial products: Permissive licenses: Whisper (tiny, base, small, medium, turbo), Qwen3-ASR 0.6B, Qwen3-ASR 1.7B, Omnilingual ASR CTC 300M, Cohere Transcribe 03-2026, VibeVoice-ASR 7B, VibeVoice-ASR Streaming 1.5B, VibeVoice-ASR-BitNet, Fun-ASR-MLT-Nano-2512, Voxtral Mini 4B Realtime 2602 and VibeVoice-ASR-Streaming-7B. 1 other model carries attribution, ShareAlike or custom terms.
CPU-only or low-power machines: Rated 4/5 or better for CPU-only use: Whisper tiny, Whisper base, Whisper small and VibeVoice-ASR-BitNet.

All 16 models that list Mandarin Chinese

Reviewed local ASR models listing Mandarin Chinese
Model	Streaming	Timestamps	Diarization	License	CPU · NVIDIA fit
Cohere Transcribe 03-2026 Cohere Labs · 2B	No	Limited	No	Apache-2.0 Permissive	1/5 · 5/5
Fun-ASR-MLT-Nano-2512 FunAudioLLM · 800M	Via runtime	No	No	Apache-2.0 Permissive	2/5 · 5/5
Nemotron 3.5 ASR Streaming 0.6B NVIDIA · 600M	Native	Limited	No	OpenMDW-1.1 Custom terms	2/5 · 5/5
Omnilingual ASR CTC 300M Meta AI · 325M	No	Limited	No	Apache-2.0 Permissive	3/5 · 5/5
Qwen3-ASR 0.6B Qwen · 0.6B	Via runtime	Separate aligner	No	Apache-2.0 Permissive	2/5 · 5/5
Qwen3-ASR 1.7B Qwen · 1.7B	Via runtime	Separate aligner	No	Apache-2.0 Permissive	1/5 · 5/5
VibeVoice-ASR 7B Microsoft · 9B total	No	Yes	Yes	MIT Permissive	1/5 · 4/5
VibeVoice-ASR Streaming 1.5B Microsoft · 3B total	Native	Limited	Yes	MIT Permissive	1/5 · 4/5
VibeVoice-ASR-BitNet Microsoft · 0.3B quantized	Via runtime	Limited	No	MIT Permissive	5/5 · 3/5
VibeVoice-ASR-Streaming-7B Microsoft · 9B total	Native	Limited	Yes	MIT Permissive	1/5 · 4/5
Voxtral Mini 4B Realtime 2602 Mistral AI · 4B	Native	Limited	No	Apache-2.0 Permissive	1/5 · 5/5
Whisper base OpenAI · 74M	Via runtime	Yes	No	MIT Permissive	5/5 · 4/5
Whisper medium OpenAI · 769M	Via runtime	Yes	No	MIT Permissive	2/5 · 5/5
Whisper small OpenAI · 244M	Via runtime	Yes	No	MIT Permissive	4/5 · 5/5
Whisper tiny OpenAI · 39M	Via runtime	Yes	No	MIT Permissive	5/5 · 4/5
Whisper turbo OpenAI · 809M	Via runtime	Yes	No	MIT Permissive	2/5 · 5/5

Multilingual models that do not list Mandarin Chinese

Parakeet TDT 0.6B v3, Orukeet, Granite Speech 5.0 2B, Granite Speech 5.0 470M TurboCTC, Parakeet Redux, Whistle and Jais-ASR are multilingual but do not list Mandarin Chinese in the reviewed official source, so the matcher does not recommend them for this language.

Cantonese has its own language code (yue) and a much shorter model list; see the Cantonese guide. The Qwen3-ASR model cards also list 22 Chinese dialects.

The same shortlist applies to Korean

The reviewed official sources list exactly the same 16 models for Korean. The rankings, requirement picks and license notes on this page therefore apply unchanged to that language; accuracy per language still differs, so test with your own recordings.

Common questions

What is the best local speech-to-text model for Mandarin Chinese?

There is no single winner; it depends on the workload. In the matcher, Whisper turbo leads for long recordings on a mid-range GPU, VibeVoice-ASR 7B for meetings with speaker labels and VibeVoice-ASR-BitNet for CPU-only machines. Benchmark the top two or three on your own Mandarin Chinese audio before deciding.

Is there an offline Mandarin Chinese model with real-time streaming?

Yes. Nemotron 3.5 ASR Streaming 0.6B, VibeVoice-ASR Streaming 1.5B, Voxtral Mini 4B Realtime 2602 and VibeVoice-ASR-Streaming-7B list Mandarin Chinese and have a native streaming architecture. Measure end-to-end latency on your hardware.

Does Whisper support Mandarin Chinese?

Yes. All Whisper sizes list Mandarin Chinese in the official tokenizer language list. Smaller sizes trade accuracy for speed; Whisper turbo is the usual starting point on a GPU.

Which models do not support Mandarin Chinese?

Among multilingual models, Parakeet TDT 0.6B v3, Orukeet, Granite Speech 5.0 2B, Granite Speech 5.0 470M TurboCTC, Parakeet Redux, Whistle and Jais-ASR do not list Mandarin Chinese in their reviewed sources. English-only models are also excluded.

Generated from the reviewed model database (latest review November 3, 2026). "Lists Mandarin Chinese" means the model's official source explicitly names the language; it does not guarantee equal accuracy across languages.