Step 5 Preview
Neural Network
Step 5 Preview is StepFun's flagship Mixture-of-Experts model for agentic scenarios and development tasks.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 048 576 tokens
- Max answer length
- 64 000 tokens
- Reasoning effort
- low, medium, high
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Step 5 Preview?
Step 5 Preview is StepFun's flagship sparse Mixture-of-Experts model: out of 600 billion parameters, about 27 billion are active at each step. It focuses on agentic work: the model excels in software development tasks and applied professional scenarios requiring long chains of reasoning. In BotHub, it is accessible from Russia without a VPN or foreign card; payment is in rubles based on actual token usage, and tokens do not expire. Over 250 other neural networks are available in the same window, making it easy to switch and compare answers for a single task. For production, there is a unified OpenAI-compatible API, allowing you to switch models without rewriting integrations. Companies have access to contracts, invoices, and an admin panel with limits. Use Step 5 Preview for autonomous agents, code analysis and refactoring, debugging complex services, preparing technical documentation, and analyzing work materials.Frequently asked questions about Step 5 Preview
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a StepFun text model with a reasoning mode and a context window of up to 1,000,000 tokens. It is suitable for analyzing large documents and codebases, writing and refactoring code, parsing reports, preparing analytics, and building agents with tool calling.
Up to 64,000 tokens in a single response—enough for a lengthy article, detailed technical analysis, a large code module, or extensive documentation. If the material is larger, break the task into parts: the million-token context allows you to keep the entire conversation history.
Yes, the reasoning mode is confirmed. The model breaks down the task step-by-step before providing a result—this significantly helps with math, logic, code debugging, and working with long documents where many fragments need to be cross-referenced.
Yes, both function calling and structured JSON output are supported. The model can be connected to external tools, databases, and services. In BotHub, this is done via a unified OpenAI-compatible API, so you can switch to another model without rewriting your integration.
According to our data, the model accepts text, images, and documents as input. You can send a screenshot of an interface, a diagram, a chart, or a PDF file and ask it to analyze the content, extract data, or explain what is depicted.
The model handles Russian-language queries just as well as other tasks, and the quality is best verified with your own material. In BotHub, you can launch it directly from Russia without a VPN or foreign card, with payment in rubles based on actual token usage.