StepFun · 2026-09-18 · major
Step 5 Preview — StepFun's 600B flagship for agent work
Step 5 Preview is StepFun's new flagship model for agent work, with 600 billion parameters, a 1M-token context window, and text, image and video input. It scores 44 on the Artificial Analysis Intelligence Index.
StepFun's new flagship handles coding, long documents and professional knowledge work in a 1M-token window.
Key specs
| Aa intelligence index | 44 |
|---|---|
| Output speed | 99.8 tokens/sec |
Quick facts
| Maker | StepFun |
|---|---|
| Released | 18 September 2026 |
| Parameters | 600B |
| Context window | 1M tokens in and out |
| Input types | Text, image, video |
| Model id | step-5-preview |
| Availability | StepFun API |
Pricing
| Input | $1.00 / 1M tokens |
|---|---|
| Output | $2.70 / 1M tokens |
What is it?
Step 5 Preview is StepFun's new flagship model for agent work, released on 18 September 2026 with 600 billion parameters. It reads text, images and video and writes text back. StepFun points it at software engineering, long-document work and professional knowledge tasks, and serves it through its API under the model id step-5-preview.
How does it work?
A 1M-token limit applies to both input and output, so a long codebase or a stack of documents fits in a single call without chunking. Three reasoning levels — low, medium and high — let you pick how much thinking the model does per request, on the same model id. Vision input accepts JPG, PNG, WebP and static GIF, up to 60 images per request, plus MP4, QuickTime and Matroska video.
Why does it matter?
Price is where Step 5 Preview stands out: $1.00 per million input tokens and $2.70 per million output tokens, against an Artificial Analysis Intelligence Index score of 44 that puts it 24th of the 200 models Artificial Analysis tracks. That combination gives teams a cheaper option for agent loops that burn a lot of tokens, without dropping to a small model.
Who is it for?
teams building coding and document agents
Frequently asked questions
- Is Step 5 Preview open source?
- Step 5 Preview is served through StepFun's API rather than as a weights download. StepFun's developer documentation covers API access only, and Artificial Analysis lists Step 5 Preview as a proprietary model. Teams that need to run a model on their own hardware will not be able to do that with Step 5 Preview as it stands today.
- How does Step 5 Preview rank against other models?
- Step 5 Preview scores 44 on the Artificial Analysis Intelligence Index, which places it 24th out of the 200 models Artificial Analysis tracks. The median score across that set is 24, so Step 5 Preview sits well above the middle of the field while costing $1.00 per million input tokens.
- What image and video formats does Step 5 Preview accept?
- Step 5 Preview accepts JPG and JPEG, PNG, WebP and static GIF images, with a limit of 60 images per request. For video, StepFun's documentation lists MP4, QuickTime and Matroska. Output is text only, so the model can describe or reason about a screenshot, chart or clip but cannot generate one.
- Does Step 5 Preview use more tokens than other models?
- Yes. Artificial Analysis measured Step 5 Preview generating 160 million tokens across its evaluation set, against a median of 92 million, and describes the model as very verbose. Because output tokens cost $2.70 per million, that verbosity eats into the low headline price on workloads that produce long answers.
Try it
Call model id `step-5-preview` on platform.stepfun.ai