Step AI

What is Step AI?

StepAI is a multi-modal AI assistant developed by Shanghai StepStar, offering individual users features such as dialogue-based Q&A, online searching, access to knowledge bases, image generation, voice capabilities, and the ability to deploy intelligent agents. In its early stages, this product was known to users as “Yuewen” or StepChat; currently, the official name used across all platforms is “StepAI”. The web version, the app, and the WeChat mini-program all belong to the same product ecosystem.

Step Star was founded in 2023 by Jiang Daxin. Its services include the Step series of basic models, Step Star AI applications, an open platform, as well as enterprise services. Individual memberships, the API offerings on the open platform, and the developer Step Plan are separate from one another; the quotas allocated to each cannot be used together.

Core functions

1. Multimodal dialogue

Users can input text, images, files, voice recordings, or links, allowing Step AI to handle tasks such as knowledge-based questions, summarization, writing, translation, coding, and creative work. The mobile version supports taking photos to pose questions, voice input, as well as intelligent voice or video calls; for these functions, the user must grant permission for access to the camera, photo album, and microphone.

The model is able to understand multimodal content, but it may misinterpret small text, tables, charts, audio names, and complex images. Important data should be checked item by item in the original file.

2. Online search

Once Web Search is enabled, StepAI will search the Internet and provide answers based on the results found; it is suitable for obtaining news background information, conducting product research, and compiling up-to-date data. The scope of the search results, their dates, and the quality of the sources are not entirely under control, and when dealing with topics related to medicine, law, finance, policies, or citations, it is necessary to check the original pages for verification.

3. Knowledge base Q&A

The web version offers a \"My Knowledge Base\" feature, allowing users to upload files and ask questions related to their private materials. This knowledge base is useful for reading contracts, reports, learning materials, and corporate documents; it is also possible to switch to a mode focused solely on answering questions using this knowledge base during conversations.

The response may omit tables, footnotes, scanned pages, or cross-file references. It is necessary to check the permissions of the documents before uploading and to remove any sensitive information.

Users can view or delete uploaded files in the knowledge base.

4. Image generation and editing

StepAI is capable of generating images from natural language; the underlying models Step 1X, Step 2X, and the image editing tools cover various styles and editing scenarios. The open platform supports square as well as vertical and horizontal formats, with the Step 1X series allowing the generation of one image at a time.

The images generated may contain spelling errors, distorted figures, or elements similar to existing works. Before using them for commercial purposes, it is necessary to check the requirements regarding portraits, trademarks, copyrights, and the labeling rules for AI-generated content set by the platform.

5. Step Audio Studio

Audio Studio offers services such as speech recognition, text-to-speech conversion, voice cloning, and various audio-related creation tools. The StepAudio series focuses on end-to-end speech understanding and generation, and is capable of handling aspects such as emotion, speech speed, dialects, and paralinguistic information.

Voice cloning can only be used with samples for which explicit authorization has been obtained; it must not be employed for impersonation or fraud.

6. StepClaw agent

StepClaw is an AI agent that can understand instructions and carry out tasks automatically. StepAI offers one-click deployment options for both desktop and cloud environments; some members have access to systems with 2 cores and 4GB of memory, or 4 cores and 8GB of memory. It also enables integration with WeChat, Lark, and Enterprise WeChat, allowing the intelligent assistant to operate continuously.

StepClaw is still marked as being in the testing phase. Once deployed, it will have access to messages, files, the network, and external tools; therefore, accounts with minimal permissions should be used, high-risk commands must be restricted, and the uploading of production keys and customer privacy data must be prevented. Additionally, budgets, logging capabilities, and a shutdown mechanism should be established for its continuous operation.

Model capability

The Step series covers text reasoning, vision, images, speech, and Agents. The key components of the current ecosystem include Step 3.7 Flash, Step 3.5 Flash, Step 3 multi-modal reasoning, Step 1o for vision, StepAudio, and Step 1X for image processing.

• Step 3.7/3.5 Flash：Designed for cost-effective reasoning, programming, tool invocation, and Agent tasks;
• Step 3：A multimodal reasoning model with 321 billion total parameters and 38 billion active parameters;
• Step 1o Vision：Used for image understanding and visual question answering;
• StepAudio：It covers end-to-end voice conversations, ASR, TTS, voice reasoning, and editing;
• Steps 1X and 2X:Used for text-to-image generation and image editing.

The app will update the underlying models, and the function names displayed on the website may not indicate the specific version. Developers who need reproducible results should use APIs to fix the models and create their own evaluation sets.

Price and version comparison

Step AI membership prices

The following are the promotional prices displayed on the official subscription page as of August 2026; the original prices, points, and computing resources associated with them may change.

• Try something new:9.9 yuan per week, including 250 points and 7 days of benefits;
• Getting started:39 yuan per month, including 1300 points, 2 cores, 4GB of memory, and 40GB of storage;
• Advanced:99 yuan per month, including 4,500 points and priority technical support;
• Advanced:199 yuan per month, including 10,000 points;
• Major:499 yuan per month, including 27,000 points, 4 cores and 8GB of memory, 40GB of storage, as well as dedicated support services.

Each plan includes one-click deployment of StepClaw on the desktop, as well as support via the cloud version of StepClaw and through WeChat, Lark, and Enterprise WeChat. Subscriptions may be renewed automatically; the rules regarding cancellation and remaining credits are subject to the purchase page and the membership agreement.

Pay-as-you-go pricing for open platforms

The open platform charges separately for the model’s inputs, cache hits, and outputs. The current representative prices are as follows:

• Step 3.5 Flash：The cost is 0.7 yuan for input, 0.14 yuan for caching, and 2.1 yuan per million tokens for output;
• Step 3：For short inputs, the cost is 1.5 yuan for the input and 0.3 yuan for caching, with an output cost of 4 to 8 yuan per million tokens; when the input size is between 4K and 64K, the cost is 4 yuan for the input, 0.8 yuan for caching, and 10 yuan for the output.
• Step 1o Turbo Vision：The input cost is 2.5 yuan, the caching cost is 0.5 yuan, and the output cost is 8 yuan per million tokens;
• Step 2 Mini：Input: 1 yuan, caching: 0.2 yuan, output: 2 yuan per million tokens;
• StepAudio 2.5 TTS：5.8 yuan per 10,000 characters; Step TTS 2 costs 2.8 yuan, while Mini costs 0.9 yuan.
• Speech recognition:The cost varies by model, ranging from approximately 0.15 to 2.6 yuan per hour;
• Voice replication:9.9 yuan per sound color;
• Step 1X Medium：0.1 yuan per sheet, while Image Edit 2 costs 0.02 yuan per sheet;
• Online search:0.04 yuan per transaction; file storage costs 0.5 yuan per GB per day.

In general, in Chinese, approximately 1.6 characters correspond to 1 token – this is just an estimate; the actual value is determined by the usage field of the interface. Whether a cache hit occurs, the length of the context, tool calls, and retries all affect the cost.

Step Plan

Step Plan is a subscription API designed for development tools and high-frequency Agent workflows. It uses separate interface addresses and keys, and comes with limits of 5 hours per usage and a weekly limit. These limits are not carried over from one period to another; the weekly limit is approximately four times the 5-hour limit.

After exceeding the limit, a corresponding period is required for reset.

The Step Plan and the regular top-up balance operate as two separate systems; using the Plan does not result in any deduction from the pay-as-you-go balance. The official documentation states that the promotional prices are temporary offers, and the actual prices of the packages should be shown in real time on the subscription page; old campaign prices cannot be presented as permanent commitments.

GitHub and the open-source status

StepStar possesses an official GitHub organization named stepfun-ai, which makes available a variety of models, research materials, and training frameworks. For projects such as Step 3.5 Flash, Step 3.7 Flash, Step3, Step-Audio, Step1X-Edit, Step-DeepResearch, GUI Agent, and SteptronOss, it is possible to find the source code, weights, or technical reports.

Step 3.5: The Flash and several new projects are licensed under Apache 2.0, while the GUI Agent project is licensed under MIT; however, the weights of different models, as well as their code, data, and commercial licenses, may vary, so it is necessary to check the license associated with each repository separately.

An open-source model does not mean that the StepAI account system, StepClaw cloud services, membership features, or the backend of the open platform are all open source.

Privacy and security

Step AI processes text, audio, images, files, links, operation commands, and feedback submitted by users, and it performs online searches when activated. The photo-taking, audio, and video functions on mobile devices require permissions for access to the camera, album, and microphone.

Users can manage sessions, favorites, shared links, knowledge base files, generate audio, and recreate sound colors. Before uploading organizational data, unpublished code, or personal information, it is necessary to confirm the relevant permissions.

Whether deleting records from the interface overrides backups and mandatory retention requirements is also determined by the privacy policy.

Step-by-Step AI Usage Guide

Complete a basic task.

1. Clarify the issue, time frame, location, source priority, and output format;
2. In Step AI, upload materials for which you have usage rights or enter a search query;
3. First, establish a framework using multimodal dialogue, and then supplement it with evidence through online searches.
4. It is necessary to distinguish between factual information from the source, the author’s opinions, and AI-generated conclusions.
5. Check each item for dates, numbers, the original location, and any conflicting evidence;
6. The conclusions are manually revised, the verification time is recorded, and then they are published;

Create reusable professional workflows

1. Break down complex topics into four categories of questions: background, data, comparison, and conclusions;
2. Combining multi-modal dialogue, online search, and knowledge base querying forms a fixed set of research steps.
3. Give priority to using the official website, research papers, regulatory documents, and raw data;
4. A second person is assigned to review conclusions that are considered high-risk;
5. Save queries, evidence, versions, and unresolved issues;
6. Re-run after the data changes and update the conclusions;

Which users is it suitable for?

• Individual users who need Chinese-language online Q&A, document summarization, and a knowledge base;
• Content creators who wish to use images, audio, and multi-modal elements in their work;
• Users who need to deploy StepClaw in order to connect to the office messaging platform;
• Developers who use cost-effective inference, vision, and speech APIs;
• Technical teams that research and deploy the Step open-source model locally;
• Customers who need industry models, enterprise integration, and dedicated support.

Usage restrictions

• AI may generate false facts, fabricated citations, and unsafe code;
• The internet and knowledge bases can provide context, but they cannot eliminate illusions.
• Agents and cloud-based persistent services present higher risks; manual approval is required when sending messages, executing commands, making purchases, deleting items, or publishing content.
• The official benchmarks and models are based on specific datasets and do not represent every real-world task;
• Before purchasing a membership or API, it is advisable to conduct small-scale tests using your own Chinese-language materials, long texts, images, audio files, and considering your concurrency requirements.

Frequently Asked Questions

What is the relationship between Step AI and Yuewen?

Yuewen was one of the early intelligent Q&A brands under StepStar; currently, the official website and legal agreements use the name “StepAI”, while the core services and product offerings have evolved and expanded to include StepClaw, knowledge bases, and multi-modal capabilities.

Is Step AI free?

Some features can be tried out for free, but higher credit limits, cloud-based StepClaw services, and additional computing resources require a membership. The basic membership costs 39 yuan per month, while there is also an option at 9.9 yuan per week for those who want to try it out first.

Are APIs provided?

Available. The open platform covers reasoning, vision, text, speech recognition and synthesis, voice replication, images, and online search; it can be used on a pay-as-you-go basis or through a step-based plan.

What is StepClaw?

It is an AI Agent that can be deployed on desktops or in the cloud, connected to messaging platforms, and capable of automating tasks; it is still in the testing phase, so special attention must be paid to permissions and data security.

Is StepAI open source?

The full application and cloud services are not open source, but various official projects such as Step 3.7/3.5 Flash, StepAudio, image editing tools, and research frameworks have their code or weights made available; the licensing terms for each of these projects need to be checked separately.