Step 5 Preview
Step 5 Preview is StepFun's flagship for agentic work, built on a sparse Mixture-of-Experts architecture with 27B parameters active out of 600B total. It is strong at software engineering and professional tasks, reads text, images and video, and carries a 1M-token context window at $1.00 per million input tokens.
Technical Specifications
Live data fetched from the OpenRouter model registry — always up to date.
What is Step 5 Preview?
Step 5 Preview is StepFun's flagship release and a good example of why sparse Mixture-of-Experts models have become the default at the top of the market: 600B total parameters, but only 27B active per token. That keeps inference cost far below what a dense model of the same size would need, which is how StepFun can price frontier-tier context at $1.00 per million input tokens.
Features & Capabilities
Sparse MoE Design
600B total parameters with 27B active - frontier scale, modest compute.
Software Engineering
Strong professional and engineering performance at its price point.
1M Context
Whole repositories and long reports in a single session.
Text, Image and Video Input
Accepts video as well as images and documents.
Agentic Focus
Tuned for multi-step autonomous work rather than single answers.
Frontier Context at $1
$1.00 per million input tokens for a 1M-token window.
Who Is Step 5 Preview For?
Cost-Efficient Agents
Agentic loops that would be expensive on a dense flagship.
Engineering Tasks
Software work that needs a full repository in view.
Video and Image Analysis
Video input alongside text for inspection tasks.
Professional Reporting
Long structured deliverables from large source material.
Step 5 Preview Pricing
- $1.00 per million input tokens, $2.70 per million output tokens
- Sparse MoE architecture keeps the price down despite frontier-scale parameters
- Run it from your Hoodgen free credits, no credit card required
- Higher allowances on Hoodgen AI paid plans
What Makes Step 5 Preview Different?
The architecture is the story here. Step 5 uses 600B total parameters but activates only 27B per token, so it behaves like a much larger model while costing far less to run. That is the same tradeoff that made Mixture-of-Experts standard in this generation, and StepFun has applied it at frontier scale.
The practical result is a 1M-token context window at $1.00 per million input tokens, with video input as well as text and images. For agentic workloads where the same large context gets sent repeatedly, that combination changes what is affordable rather than just what is possible.
It is labelled Preview, which is worth taking at face value: expect the final release to shift. For evaluation and cost planning it is a low-risk way to test the architecture, and Hoodgen will carry the final version when it lands.
Step 5 Preview Performance & Pricing
Step 5 Preview competes on architecture efficiency - frontier-scale context and MoE design at a price below the established flagships.
What Can You Do With Step 5 Preview?
Step 5 Preview is a sensible choice when you want a 1M-token window for agentic work without paying flagship prices for it. Video input adds a second use case that cheaper models do not cover.
Best Use Cases
- Agentic coding with a large repository in context
- Video inspection alongside written requirements
- Long professional reports from source material
- Evaluation work ahead of the final release
- High-volume pipelines that need long context
Step 5 Preview Privacy & Data Policy
Step 5 Preview is developed by StepFun and served through OpenRouter. Prompts and completions are processed under StepFun terms.
As with any third-party model, avoid sending credentials or personal data. Hoodgen AI keeps other long-context models on the same account so you can compare them without rebuilding your prompts.
Step 5 Preview vs Other Models
Frequently Asked Questions
Frontier context at a fraction of the price
Chat with Step 5 Preview on Hoodgen - sign in and start in seconds.
Chat with Step 5 Preview