Artificial Analysis has launched version 2.0 of its AA-WER speech-to-text benchmark, showcasing notable advancements in AI transcription technology. Leading the pack is ElevenLabs’ Scribe v2, boasting a remarkable word error rate (WER) of just 2.3%. Close competitors include Google’s Gemini 3 Pro with a WER of 2.9% and Mistral’s Voxtral Small at 3.0%. Google’s Gemini 3 Flash (3.1%) and ElevenLabs’ older Scribe v1 (3.2%) follow closely behind. Notably, OpenAI’s Whisper Large v3 achieved a WER of 4.2%. In the specialized AA-AgentTalk test, focused on voice assistant interactions, Scribe v2 (1.6%) and Gemini 3 Pro (1.7%) maintained their leading positions. AssemblyAI’s Universal-3 Pro completed the top three with a WER of 2.3%. The AA-WER 2.0 results highlight significant innovations in speech recognition, with the top performers emphasizing the efficiency and accuracy of AI-driven transcription solutions.
Source link
