← all releases
StepFun AI·Nov 29, 2025·v1.0·open source

Step-Audio-R1

Step-Audio-R1 is an advanced audio language model developed by StepFun AI, designed to enhance audio reasoning capabilities by grounding its reasoning in acoustic features. It introduces Modality-Grounded Reasoning Distillation (MGRD), an iterative training framework that shifts the model's reasoning from textual abstractions to acoustic properties, effectively addressing the 'inverted scaling' problem where performance degrades with longer reasoning. This model has demonstrated superior performance across various audio understanding and reasoning benchmarks, surpassing models like Gemini 2.5 Pro and achieving results comparable to Gemini 3 Pro.

LLMsMLTesting
visit website
overview

Step-Audio-R1 is an advanced audio language model developed by StepFun AI, designed to enhance audio reasoning capabilities by grounding its reasoning in acoustic features. It introduces Modality-Grounded Reasoning Distillation (MGRD), an iterative training framework that shifts the model's reasoning from textual abstractions to acoustic properties, effectively addressing the 'inverted scaling' problem where performance degrades with longer reasoning. This model has demonstrated superior performance across various audio understanding and reasoning benchmarks, surpassing models like Gemini 2.5 Pro and achieving results comparable to Gemini 3 Pro.

key features
  • 01Chain-of-Thought (CoT) Reasoning: Unlocks CoT reasoning capabilities in audio language models, generating reasoning chains grounded in acoustic features.
  • 02Modality-Grounded Reasoning Distillation (MGRD): An iterative training framework that shifts reasoning from textual abstractions to acoustic properties.
  • 03Superior Performance: Surpasses Gemini 2.5 Pro and is comparable to Gemini 3 across major audio reasoning tasks.
  • 04Versatility: Covers speech, environmental sounds, and music domains.
  • 05Open Source: Released under the Apache 2.0 license, promoting accessibility and collaboration.
related productsbrowse all →
SINQopen sourceSINQ (Sinkhorn-Normalized Quantization) is a novel, fast, and high-quality quantization method designed to make any Large Language Model (LLM) smaller while preserving accuracy. It offers a plug-and-play, model-agnostic technique that delivers state-of-the-art performance for LLMs without sacrificing accuracy.AIBuildAIopen sourceAIBuildAI is an AI agent that autonomously constructs AI models. Given a specific task, it initiates an agent loop to analyze the problem, design models, and execute training processes, all without human intervention.DetectFlowopen sourceDetectFlow is an open-source cybersecurity detection platform developed by SOC Prime. It leverages artificial intelligence to enhance the detection of cyber threats in real-time, enabling security operations teams to identify and respond to attacks more effectively.