Kimi-Audio
Kimi-Audio is a language-capable audio foundation-model family for audio understanding, text responses, and speech generation. It combines semantic and acoustic audio representations, pretraining on more than 13 million hours of speech, music, sounds, and paired text. The released 7B-base-language-model and instruction-tuned checkpoints support tasks including audio question answering, speech recognition, and spoken conversation. Because it generates and reasons with language, it is in scope beyond speech-only transcription or synthesis. Public MIT-tagged weights and inference code are available.
2025-04-26
7B language backbone
Audio-language Transformer built from Qwen2.5-7B with audio encoders and speech decoder
MIT
Specifications
- Parameters
- 7B language backbone
- Architecture
- Audio-language Transformer built from Qwen2.5-7B with audio encoders and speech decoder
- License
- MIT
- Type
- multimodal
- Modalities
- textaudio
Benchmark Scores
Advanced Specifications
- Model Family
- Kimi
- Finetuned From
- Qwen2.5-7B
- API Access
- Not Available
- Chat Interface
- Not Available
- Variants
- Kimi-Audio-7BKimi-Audio-7B-Instruct
Capabilities & Limitations
- Capabilities
- audio question answeringspeech understandingspoken conversationspeech generation
- Known Limitations
- The 7B name refers to the language backbone; additional audio components increase total size.Qwen-derived code retains Apache-2.0 terms alongside MIT-licensed components.
- Notable Use Cases
- audio assistantssound and music understandingspeech-language research