语音转文本 AI 模型与服务商排行榜

比较不同语音转文本模型和服务商的词错误率、速度与价格。

有关更多详细信息，请参阅我们的方法论页面。

查看流式基准测试

文本转语音竞技场

AI 语音探索器

WER 指数（非流式）

AA-WER v2 · % of words transcribed incorrectly · Lower is better

速度因子

Input audio seconds transcribed per second · Higher is better

价格

USD per 1000 minutes of audio · Lower is better

Artificial Analysis 词错误率指数（非流式）

% of words transcribed incorrectly · Lower is better · AA-WER v2 incorporates 3 datasets: AA-AgentTalk (50%), VoxPopuli-Cleaned-AA (25%), Earnings22-Cleaned-AA (25%)

注意： 对于 Earnings22，如果模型因时间限制而无法可靠处理完整长度的音频，我们会将音频切分为约 9 分钟的片段（相关模型：GPT-4o Mini Transcribe, OpenAINova 2 Pro, AmazonGPT-4o Transcribe, OpenAI）。对于时间限制更短的模型，我们会将音频切分为约 30 秒的片段（相关模型：Inworld STT 1Canary Qwen 2.5B, NVIDIAQwen3 ASR Flash, Alibaba）。

Measures transcription accuracy across 3 datasets to evaluate models in real-world speech with diverse accents, domain-specific language, and challenging channel & acoustic conditions.

AA-WER is calculated as an audio-duration-weighted average of WER across ~8 hours from three datasets: AA-AgentTalk (50%), VoxPopuli-Cleaned-AA (25%), and Earnings22-Cleaned-AA (25%). See methodology for more detail.

AA-WER（非流式）按数据集

AA-WER（非流式）：AA-AgentTalk 数据集

% of words transcribed incorrectly on the AA-AgentTalk dataset · Lower is better

Measures transcription accuracy across 3 datasets to evaluate models in real-world speech with diverse accents, domain-specific language, and challenging channel & acoustic conditions.

清洗数据集对比

VoxPopuli：公开数据的清洗子集 vs. 原始子集

% WER (word error rate) · Lower is better

排序依据

注意：清洗版本会从参考文本中移除转写错误，从而为模型评估提供更准确的基准真值。

Measures transcription accuracy across 3 datasets to evaluate models in real-world speech with diverse accents, domain-specific language, and challenging channel & acoustic conditions.

API 基准测试

Artificial Analysis 词错误率指数（非流式）vs. 价格

% of words transcribed incorrectly · Lower is better · AA-WER v2 incorporates 3 datasets: AA-AgentTalk (50%), VoxPopuli-Cleaned-AA (25%), Earnings22-Cleaned-AA (25%) · USD per 1000 minutes of audio

Most attractive quadrant

Measures transcription accuracy across 3 datasets to evaluate models in real-world speech with diverse accents, domain-specific language, and challenging channel & acoustic conditions.

Estimated cost in USD to transcribe 1,000 minutes of audio, normalized across providers with different billing models, and including billed reasoning tokens where available. Further detail on the methodology page.

速度因子

Input audio seconds transcribed per second · Higher is better

Audio file seconds transcribed per second of processing time. Higher factor indicates faster transcription speed. Reported Speed Factor values are medians across benchmark trials from the last 7 days; over-time chart points are daily medians. Artificial Analysis measurements are based on an audio duration of 10 minutes. Speed Factor may vary for other durations, particularly very short durations under 1 minute.

转写价格

USD per 1000 minutes of audio

关键指标摘要与更多信息

	服务商	Whisper 版本				更多详情
Qwen3.5 Omni Flash	Alibaba Cloud		13.5%	73.9	0.00	详情
Qwen3.5 Omni Plus	Alibaba Cloud		3.5%	173.0	0.00	详情
Fun-Realtime-ASR-preview	Alibaba Cloud		1.7%	21.0	0.00	详情
Nova 2 Pro	Amazon Bedrock		4.9%	24.0	3.10	详情
Amazon Transcribe	Amazon Bedrock		4.1%	28.0	6.00	详情
Universal-3 Pro	AssemblyAI		3.1%	101.1	3.50	详情
Universal, AssemblyAI	AssemblyAI		3.8%	118.5	2.50	详情
MAI-Transcribe-2	Microsoft AI		2.0%	303.6	1.67	详情
MAI-Transcribe-1.5	Microsoft AI		2.4%	193.6	6.00	详情
transcribe-03-2026	Cohere		4.6%	124.0	0.00	详情
Nova-3	Deepgram		5.2%	621.8	4.30	详情
Nova-3, Telnyx	Telnyx		4.8%	401.7	7.40	详情
Nova-2, Telnyx	Telnyx		5.1%	471.1	7.40	详情
Scribe v2	ElevenLabs		2.2%	76.5	3.67	详情
Solaria-1, Gladia	Gladia		4.1%	73.0	10.17	详情
Solaria-3, Gladia	Gladia		3.2%	65.7	10.16	详情
Chirp 3, Google	Google		4.3%	28.7	16.00	详情
Chirp	Google		31.2%	14.3	16.00	详情
Gemini 3.5 Transcribe	Google		2.6%	85.9	5.00	详情
Gemini 3.1 Pro Preview (High)	Google		2.8%	5.9	18.15	详情
Gemini 3.1 Pro Preview (Low)	Google		3.6%	9.3	7.72	详情
Gemini 3 Flash (High)	Google		2.9%	20.4	13.70	详情
Gemini 2.5 Flash Lite	Google		5.2%	79.7	6.56	详情
Gemini 2.5 Flash	Google		5.1%	67.5	6.66	详情
Gemini 2.5 Pro	Google		2.9%	11.6	11.39	详情
Gemini 3.1 Flash-Lite Preview (Minimal)	Google		3.4%	76.0	5.83	详情
Gradium Speech-to-Text	Gradium		6.8%	2.3	13.00	详情
Grok Voice Transcribe 2.0	SpaceXAI		2.3%	172.3	1.67	详情
Grok Voice Transcribe 1.0	SpaceXAI		4.0%	154.6	1.67	详情
Inworld STT 1	Inworld		3.9%	120.4	2.50	详情
Voxtral Mini Transcribe 2	Mistral		3.6%	86.5	3.00	详情
Voxtral Small	Mistral		2.8%	65.6	4.00	详情
Voxtral Mini	DeepInfra		3.8%	51.3	1.00	详情
Modulate STT Batch English VFast	Modulate		4.2%	59.2	0.42	详情
Canary Qwen 2.5B, NVIDIA	Replicate		4.3%	8.8	0.74	详情
Parakeet TDT 0.6B V2, NVIDIA	NVIDIA		6.4%	92.5	0.00	详情
Parakeet RNNT 1.1B	Replicate		5.4%	6.3	1.91	详情
GPT Transcribe, OpenAI	OpenAI		3.3%	45.1	4.50	详情
GPT-4o Transcribe	OpenAI		4.0%	36.0	6.00	详情
GPT-4o Mini Transcribe	OpenAI		4.5%	44.7	3.00	详情
Smallest AI Pulse Pro	Smallest.ai		2.4%	244.2	4.00	详情
Resonant-1	Reson8		3.4%	324.5	3.60	详情
Rev AI	Rev AI		5.9%	12.3	3.33	详情
Smallest AI Pulse	Smallest.ai		4.4%	260.8	5.00	详情
Soniox v5 Async	Soniox		3.8%	35.6	1.66	详情
Speechmatics Melia	Speechmatics		4.9%	165.9	4.00	详情
Speechmatics Standard	Speechmatics		5.1%	91.7	7.50	详情
Speechmatics Enhanced	Speechmatics		4.0%	44.2	12.50	详情
StepAudio 2.5 ASR, StepFun	StepFun		4.7%	116.0	0.37	详情
Whisper Large v3 Turbo	Groq	v3 Turbo	4.6%	186.7	0.67	详情
Whisper Large v3 Turbo, Telnyx	Telnyx		5.5%	39.2	15.00	详情
Wizper Large v3	fal.ai	large-v3	4.7%	163.4	0.50	详情
Incredibly Fast Whisper	Replicate	large-v3	5.7%	57.0	1.49	详情
Whisper Large v3	Replicate	large-v3	10.1%	2.7	4.23	详情
Whisper Large v3	fal.ai	large-v3	4.1%	75.1	1.15	详情
Whisper Large v2	OpenAI	large-v2	4.1%	26.8	6.00	详情

常见问题

StepAudio 3 ASR 以最低的 AA-WER（Artificial Analysis 词错误率）1.7% 在 57 个已评估模型中领先。

按准确度（AA-WER）排名的顶级语音转文本模型为：1. StepAudio 3 ASR（1.7%）、2. Fun-Realtime-ASR-preview（1.7%）、3. MAI-Transcribe-2（2.0%）、4. Scribe v2, ElevenLabs（2.2%）和5. Grok Voice Transcribe 2.0（2.3%）。AA-WER 越低，表示转写准确度越高。

Nova-3 最快，速度因子为实时的 621.8x，其次是 Nova-2, Telnyx（471.1x）和 Nova-3, Telnyx（401.7x）。速度因子越高，转写越快。

StepAudio 2.5 ASR 最实惠，价格为每 1,000 分钟 $0.3667，其次是 Modulate STT Batch English VFast（$0.417）和 Wizper (L, v3), fal.ai（$0.50）。

Voxtral Small, Mistral 是最准确的开放权重模型，AA-WER 为 2.8%。在总共 57 个已评估模型中，有 13 个开放权重模型。

按准确度排名的顶级开放权重语音转文本模型为：1. Voxtral Small, Mistral（AA-WER 2.8%）、2. Inkling (256K), Thinking Machines（AA-WER 3.5%）和3. Voxtral Mini Transcribe 2, Mistral（AA-WER 3.6%）。

最佳模型取决于你的优先级。使用散点图可视化准确度（AA-WER）、速度与价格之间的权衡。对于需要高准确度的应用，优先选择 AA-WER 更低的模型；对于实时应用，关注速度因子；对于成本敏感的工作负载，比较价格图表。

语音转文本 AI 模型与服务商排行榜

相关链接

WER 指数（非流式）

速度因子

价格

Artificial Analysis 词错误率指数（非流式）

Artificial Analysis 词错误率指数（非流式）

Artificial Analysis Word Error Rate (AA-WER) Index

AA-WER（非流式）按数据集

AA-WER（非流式）：AA-AgentTalk 数据集

Artificial Analysis Word Error Rate (AA-WER) Index

清洗数据集对比

VoxPopuli：公开数据的清洗子集 vs. 原始子集

Artificial Analysis Word Error Rate (AA-WER) Index

API 基准测试

Artificial Analysis 词错误率指数（非流式）vs. 价格

Artificial Analysis Word Error Rate (AA-WER) Index

Price

速度因子

Speed Factor

转写价格

Price

关键指标摘要与更多信息

常见问题

哪个语音转文本模型最准确？

顶级语音转文本模型有哪些？

哪个语音转文本模型最快？

哪个语音转文本模型最便宜？

哪个开放权重语音转文本模型最好？

顶级开放权重语音转文本模型有哪些？

如何选择最佳语音转文本模型？