StepFun Open Platform Documentation

StepFun’s flagship model for agentic work. It delivers frontier-level performance across software engineering and professional knowledge work, with particular strength in finance. It natively supports text, image, and video input and can use tools and documents to carry multi-step tasks through to completion. API model ID: step-5-preview.

StepFun’s flagship multimodal reasoning model. Building on Step 3.5 Flash’s high-speed reasoning and tool-calling capabilities, it adds native multimodal input and can understand images and videos without an additional vision MCP or auxiliary model. It supports low, medium, and high reasoning effort.

A flagship reasoning model built for agents. It combines deep reasoning with fast responses and stable, reliable tool calling, and is particularly effective for complex project planning and long-horizon task execution.

Designed for social content creation, AI assistants, and image and video understanding. A single request accepts up to 60 images with a combined size of up to 20 MB, or an MP4 video smaller than 128 MB.

A lightweight model that supports both text-to-image generation and image editing. A single edit completes in 1-2 seconds, and input images can be up to 4096 x 4096 pixels.

Generates realistic images with strong Chinese and English text rendering. Text-to-image prompts can contain up to 512 characters and image-to-image prompts up to 1,024 characters. Input images must be PNG or JPEG, no larger than 10 MB or 2048 x 2048 pixels.

A full-duplex voice flagship for realtime interaction. It combines realtime audio understanding, natural interruption handling, turn coordination, adaptive reasoning, think-while-speaking behavior, and Voice Agent tool calling for in-vehicle systems, smart devices, customer service, and companion applications.

A conversational model designed for natural, human-like interaction with text output. It understands complex meaning and paralinguistic cues such as hesitation and laughter, produces emotionally aware responses, and supports fine-grained persona customization.

Supports Mandarin Chinese, English, Japanese, Cantonese, and Sichuanese with 19 official voices. It also supports high-quality voice cloning in Chinese, English, and Japanese for customer outreach, companion applications, and voice assistants that require natural speech.

A next-generation streaming speech recognition model based on a 4B MTP architecture. It balances recognition accuracy and response latency, supports Chinese and English with ITN text normalization, and is designed for realtime captions, voice input, and meeting transcription.

A 32B-parameter professional automatic speech recognition model.

An ASR model with strong Chinese and English recognition. It distinguishes speech from noise and supports mixed Chinese-English speech and heavily accented Mandarin for voice input, voice control, and meeting transcription.

Provides ultra-low-latency, bidirectional voice conversation. It accepts Chinese, English, and heavily accented Mandarin, and can respond in Chinese, English, Japanese, Cantonese, and Sichuanese. A single interaction can last up to 30 minutes and process up to 70 minutes of audio.

StepFun’s flagship model for agentic work. It delivers frontier-level performance across software engineering and professional knowledge work, with particular strength in finance. It natively supports text, image, and video input and can use tools and documents to carry multi-step tasks through to completion. API model ID: step-5-preview.

StepFun’s flagship multimodal reasoning model. Building on Step 3.5 Flash’s high-speed reasoning and tool-calling capabilities, it adds native multimodal input and can understand images and videos without an additional vision MCP or auxiliary model. It supports low, medium, and high reasoning effort.

A flagship reasoning model built for agents. It combines deep reasoning with fast responses and stable, reliable tool calling, and is particularly effective for complex project planning and long-horizon task execution.

StepFun’s flagship model for agentic work. It delivers frontier-level performance across software engineering and professional knowledge work, with particular strength in finance. It natively supports text, image, and video input and can use tools and documents to carry multi-step tasks through to completion. API model ID: step-5-preview.

StepFun’s flagship multimodal reasoning model. Building on Step 3.5 Flash’s high-speed reasoning and tool-calling capabilities, it adds native multimodal input and can understand images and videos without an additional vision MCP or auxiliary model. It supports low, medium, and high reasoning effort.

Designed for social content creation, AI assistants, and image and video understanding. A single request accepts up to 60 images with a combined size of up to 20 MB, or an MP4 video smaller than 128 MB.

A full-duplex voice flagship for realtime interaction. It combines realtime audio understanding, natural interruption handling, turn coordination, adaptive reasoning, think-while-speaking behavior, and Voice Agent tool calling for in-vehicle systems, smart devices, customer service, and companion applications.

A conversational model designed for natural, human-like interaction with text output. It understands complex meaning and paralinguistic cues such as hesitation and laughter, produces emotionally aware responses, and supports fine-grained persona customization.

The model closely follows natural human speech in timbre, intonation, rhythm, breathing, and pauses. It can express laughter, hesitation, stutters, repetition, and self-correction while adapting emotion and tone to the meaning of the text.

Supports Mandarin Chinese, English, Japanese, Cantonese, and Sichuanese with 19 official voices. It also supports high-quality voice cloning in Chinese, English, and Japanese for customer outreach, companion applications, and voice assistants that require natural speech.

A next-generation streaming speech recognition model based on a 4B MTP architecture. It balances recognition accuracy and response latency, supports Chinese and English with ITN text normalization, and is designed for realtime captions, voice input, and meeting transcription.

A 32B-parameter professional automatic speech recognition model.

An ASR model with strong Chinese and English recognition. It distinguishes speech from noise and supports mixed Chinese-English speech and heavily accented Mandarin for voice input, voice control, and meeting transcription.

Provides ultra-low-latency, bidirectional voice conversation. It accepts Chinese, English, and heavily accented Mandarin, and can respond in Chinese, English, Japanese, Cantonese, and Sichuanese. A single interaction can last up to 30 minutes and process up to 70 minutes of audio.