Did Codex Reset
GitHub

StepFun releases StepAudio 3 ASR at 1.7% WER on the AA-WER Index

Artificial Analysis

Artificial Analysis says StepFun has released StepAudio 3 ASR, a speech-to-text model available through the StepFun API for non-streaming transcription. It is the first StepFun model to reach the top of Artificial Analysis’s non-streaming speech-to-text leaderboard, with a 1.7% word error rate on the AA-WER Index, effectively tied with Alibaba’s Fun-Realtime-ASR-preview at 1.7%. Artificial Analysis recorded 4.7% WER for StepAudio 2.5 ASR.

On AA-AgentTalk, Artificial Analysis’s held-out voice-agent dataset, StepAudio 3 ASR scores 1.4% WER. On long-form Earnings22 calls it scores 2.8%, behind Fun-Realtime-ASR-preview at 1.8%.

Artificial Analysis measures its transcription speed at 88x real time, behind MAI-Transcribe-2 at 374x, Smallest AI Pulse Pro at 285x and Grok Voice Transcribe 2.0 at 154x. That is roughly level with ElevenLabs Scribe v2 at 84x and Gemini 3.5 Transcribe at 91x, and ahead of Fun-Realtime-ASR-preview at 20x. StepAudio 3 ASR costs $0.40 per hour, or $6.67 per 1,000 minutes. Artificial Analysis calls it the most expensive of the five most accurate models; MAI-Transcribe-2 and Grok Voice Transcribe 2.0 cost $1.67 per 1,000 minutes, and ElevenLabs Scribe v2 costs $3.67 per 1,000 minutes.

#stepfun#stepaudio#speech-to-text#artificial-analysis#asr

Artificial Analysis ranks Ming-Image-0.1-Design first among open-weight UI/UX models

On the Artificial Analysis Text to Image Leaderboard, the 6B Ming-Image-0.1-Design model places 17th of 81 in UI/UX design and first among open-weight models, compared with 45th of 160 overall. It is released under the MIT license with RGBA output and was evaluated at 2K resolution using 12 inference steps.