Model Overview
Kimi-Audio-7B-Instruct is an instruction-tuned version of the Kimi-Audio audio foundation model, designed for audio understanding, generation, and dialogue. The model weights and inference code for this version were officially released on April 25, 2025. (github.com)
Capabilities and Uses
- Supports speech recognition, audio question answering, audio content description, speech emotion recognition, and sound event and scene classification.
- Supports end-to-end speech dialogue, generating text or speech and text responses from audio input; official examples also demonstrate multi-turn audio dialogue.
- Can be used for speech transcription, audio content analysis, speech interaction prototyping, and audio model research. (huggingface.co)
Limitations
- Audio recognition and generation results may be incorrect; human verification is recommended in important scenarios.
- Self-deployment requires configuring the inference environment and compute resources; the official release provides model weights, code, and deployment examples. (huggingface.co)
- The availability of this specific version on the official Moonshot AI API, its API model ID, context window, and audio duration limit have not been verified.
Pricing
As of the November 18, 2026 check, no official RMB or USD API pricing for this specific model had been verified, so the billing amounts and units for input and output remain unverified. The official domestic and international pricing pages list other models; their prices should not be applied to this model. Self-deployment also incurs compute and operating costs. (platform.kimi.com)