StepAudio：API 服务商基准测试与分析
分析 StepAudio 的 API 服务商在 Artificial Analysis 词错误率指数、速度与价格等性能指标上的表现。
亮点
Artificial Analysis 词错误率（AA-WER）指数（按 API）
Artificial Analysis 词错误率（AA-WER）指数（按 API）
% 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：Nova 2 Pro, AmazonGPT-4o Transcribe, OpenAIGPT-4o Mini 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：StepAudio 2.5 ASR, StepFunCanary Qwen 2.5B, NVIDIAInworld STT 1）。
API 基准测试
Artificial Analysis 词错误率指数 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
速度因子
速度因子
Input audio seconds transcribed per second · Higher is better
价格
转写价格
USD per 1000 minutes of audio
关键指标摘要与更多信息
服务商 | 更多详情 | ||||
|---|---|---|---|---|---|
Qwen3.5 Omni Flash |  Alibaba Cloud | 13.5% | 77.7 | 0.00 | |
Qwen3.5 Omni Plus | Alibaba Cloud | 3.5% | 95.4 | 0.00 | |
Nova 2 Pro |  Amazon Bedrock | 4.9% | 22.9 | 3.10 | |
Amazon Transcribe | Amazon Bedrock | 4.1% | 18.5 | 6.00 | |
Universal-3 Pro |  AssemblyAI | 3.1% | 100.3 | 3.50 | |
Universal, AssemblyAI | AssemblyAI | 3.8% | 123.3 | 2.50 | |
MAI-Transcribe-1.5 |  Microsoft Azure | 2.4% | 189.8 | 6.00 | |
MAI-Transcribe-1 | Microsoft Azure | 2.6% | 67.6 | 6.00 | |
transcribe-03-2026 |  Cohere | 4.6% | 119.1 | 0.00 | |
Nova-3 |  Deepgram | 5.2% | 545.1 | 4.30 | |
Scribe v2 |  ElevenLabs | 2.2% | 51.7 | 3.67 | |
Solaria-1, Gladia |  Gladia | 4.1% | 81.0 | 10.17 | |
Solaria-3, Gladia | Gladia | 3.2% | 63.5 | 10.16 | |
Gemini 3.5 Transcribe |  Google | 2.6% | 85.5 | 5.00 | |
Gemini 3.1 Pro Preview (High) | Google | 2.8% | 7.3 | 18.15 | |
Gemini 3.1 Pro Preview (Low) | Google | 3.6% | 6.7 | 7.72 | |
Gemini 3 Flash (High) | Google | 2.9% | 18.7 | 13.70 | |
Gemini 2.5 Flash Lite | Google | 5.2% | 83.3 | 6.56 | |
Gemini 2.5 Flash | Google | 5.1% | 78.3 | 6.66 | |
Gemini 2.5 Pro | Google | 2.9% | 12.3 | 11.39 | |
Gemini 3.1 Flash-Lite Preview (Minimal) | Google | 3.4% | 80.8 | 5.83 | |
Gradium Speech-to-Text |  Gradium | 6.8% | 2.3 | 13.00 | |
Grok Speech to Text, SpaceXAI |  SpaceXAI | 4.0% | 225.0 | 1.67 | |
Inworld STT 1 |  Inworld | 3.9% | 207.8 | 2.50 | |
Voxtral Mini Transcribe 2 |  Mistral | 3.6% | 81.6 | 3.00 | |
Voxtral Small | Mistral | 2.8% | 66.3 | 4.00 | |
Voxtral Mini |  DeepInfra | 3.8% | 79.1 | 1.00 | |
Modulate STT Batch English VFast |  Modulate | 4.2% | 61.9 | 0.42 | |
Parakeet TDT 0.6B V3, Togetherai |  Together AI | 4.5% | 284.4 | 1.50 | |
Canary Qwen 2.5B, NVIDIA |  Replicate | 4.3% | 7.1 | 0.74 | |
Parakeet TDT 0.6B V2, NVIDIA |  NVIDIA | 6.4% | 100.5 | 0.00 | |
Parakeet RNNT 1.1B | Replicate | 5.4% | 6.3 | 1.91 | |
GPT Transcribe, OpenAI |  OpenAI | 3.3% | 40.1 | 4.50 | |
GPT-4o Transcribe | OpenAI | 4.0% | 37.0 | 6.00 | |
GPT-4o Mini Transcribe | OpenAI | 4.5% | 41.8 | 3.00 | |
Smallest AI Pulse Pro |  Smallest.ai | 2.4% | 272.4 | 4.00 | |
Resonant-1 |  Reson8 | 3.4% | 333.1 | 3.60 | |
Rev AI |  Rev AI | 5.9% | 12.7 | 3.33 | |
Smallest AI Pulse | Smallest.ai | 4.4% | 278.3 | 5.00 | |
Soniox v5 Async |  Soniox | 3.8% | 36.8 | 1.66 | |
Soniox V4 | Soniox | 3.9% | 40.8 | 1.66 | |
Speechmatics Melia |  Speechmatics | 4.9% | 204.2 | 4.00 | |
Speechmatics Standard | Speechmatics | 5.1% | 105.6 | 7.50 | |
Speechmatics Enhanced | Speechmatics | 4.0% | 70.6 | 12.50 | |
StepAudio 2.5 ASR, StepFun | StepFun | 4.7% | 82.5 | 0.37 | |
Whisper Large v3 Turbo |  Groq | 4.6% | 115.7 | 0.67 | |
Wizper Large v3 |  fal.ai | 4.7% | 281.5 | 0.50 | |
Incredibly Fast Whisper | Replicate | 5.7% | 55.0 | 1.49 | |
Whisper Large v3 | Replicate | 10.1% | 2.5 | 4.23 | |
Whisper Large v3 | fal.ai | 4.1% | 74.7 | 1.15 | |
Whisper Large v3 | Together AI | 4.5% | 289.6 | 1.50 | |
Whisper Large v2 | OpenAI | 4.1% | 29.3 | 6.00 |