Overview
An end-to-end audio-understanding and speech-conversation release group spanning audio semantics, paralinguistic information, tool use, and multimodal retrieval.
Step-Audio 2
An end-to-end audio-understanding and speech-conversation release group spanning audio semantics, paralinguistic information, tool use, and multimodal retrieval.
MODEL NUMBER343
- First public date
- 2025-07-23
- Organization
- StepFun
- Family
- StepAudio
01
Specifications
Only published specifications are shown.
This record covers multiple tiers. Modalities and capabilities describe the group coverage; support and licensing depend on the individual tier.
- Input modalities
- Text · Audio
- Output modalities
- Text · Audio
- Open weights
- Yes
- License
- Apache-2.0
- Status
- Current
Step-Audio 2 mini · Apache-2.0 · in: Text/Audio · out: Text/AudioStep-Audio 2 mini Base · Apache-2.0 · in: Text/Audio · out: Text/AudioStep-Audio 2 mini Think · Apache-2.0 · in: Text/Audio · out: Text/Audio
02
Overview
Why it mattered
Step-Audio 2 moved StepFun’s speech capability into reproducible open weights and inference code.
03
Architecture & capabilities
Model capabilities
Native audioReasoningTool use
04
Release history
Step-Audio 2 technical report and audio-model line were introduced.
View sourceStep-Audio 2 mini and mini Base open weights were released.
View sourceStep-Audio 2 mini Think was released.
View source05
Related models
Select a card to view the related model.
No related models recorded yet
06
Sources
Suggest a correction
Copy this template, add the field, proposed value, and primary source, and share it with the archive maintainer.