Community Articles
via Decrypt · By Decrypt Editorial
StepFun's Voice AI Topped Every Benchmark. It Also Hears Your Sighs
DE
Decrypt Editorial(03:29 PM UTC)
1 min read
In brief
- StepAudio 2.5 Realtime is an end-to-end real-time speech model with fully customizable personas in Chinese and English.
- StepFun claims first place across all five voice AI benchmarks tested in April 2026, beating GPT Realtime 1.5 and Gemini Live.
- The model was trained on a million-scale persona dataset and tuned with roleplay-specific RLHF to fix a failure mode most voice AI still can't…
Add COINOTAG as a Preferred Source
Add COINOTAG to your preferred sources in Google News and Search to see our coverage first.
Add on GoogleComments
Comments
Other Community Articles