StepFun
Step 3.7 Flash
Total Context
256K
Max Output
256K
Released
N/A
step-5-previewStep 5 Preview is StepFun’s flagship model for complex reasoning, software engineering, and multi-step agent workflows. With native support for text, image, and video input and a 1-million-token context window, it can analyze extensive documents, codebases, charts, and visual content within a single request. Built for practical problem-solving, the model supports tool calling, structured outputs, and adjustable reasoning effort. Its strengths span coding, research, and professional knowledge work, with particular emphasis on financial analysis. Step 5 Preview is a strong choice for applications that need to combine information from multiple sources and work toward detailed, actionable results.
Context Window
1M tokens
Maximum Output
1M tokens
Release Date
Sep 16, 2026
Modalities
| Input Price | Output Price | Cache Read |
|---|---|---|
| $1/M | $2.7/M | $0.05/M |
Reasoning
Tool calling
Temperature parameter
Attachments
Knowledge Base
Endpoint Protocols
Verified strengths and model characteristics.
A 600B-parameter MoE activates 27B parameters per token, limiting active computation.
A 1M-token context window and image input support work across extensive text and visual material.
Official demonstrations show iterative engineering that uses prior results and execution feedback to guide further work.
Practical workloads supported by official examples.
Implement features, fix bugs, and refine interactive applications through repeated execution and feedback.
Collect and reconcile sources, then produce analytical workbooks or reports with traceable evidence.
Reconcile financial reports and draft valuation analyses with explicit assumptions and reproducible calculations for expert review.
import OpenAI from "openai"
const client = new OpenAI({
apiKey: process.env.TOKENHUB_API_KEY,
baseURL: "https://us-api.tokenhub.com/v1",
})
const result = await client.chat.completions.create({})
console.log(result.choices[0]?.message?.content)Step 5 Preview
| Index score | ||
|---|---|---|
| Artificial Analysis Intelligence Index | Artificial Analysis broad capability aggregate | 43.7 |
Metrics sourced from Artificial Analysis
Model selection, limitations, and access.
StepFun’s flagship preview model combines sparse MoE architecture, a 1M-token context, and vision for coding and professional work.
Consider software changes, research reports, analytical workbooks, and financial analysis requiring multiple rounds of evidence gathering and refinement.
StepFun reports remaining weaknesses on its hardest, longest engineering tasks. Compare task success, latency, and total token cost on representative workloads.
Check that your TokenHub account lists the model, then use its documented model ID, endpoint, and credentials. Provider availability does not confirm TokenHub routing.
As of October 9, 2026, StepFun offers product and API access and schedules open weights for October 15. Check TokenHub for its own current rates and availability.
Use one API key to access Step 5 Preview and more AI models through TokenHub.
Step 5 Preview Media and Demos
Verified public announcements, demonstrations, and community discussions.
X (Twitter)
Step 5 Preview: Official Launch
StepFun
View originalStep 5 Preview: OpenRouter And Coding Integrations
StepFun
View originalStep 5 Preview: OpenCode Trial Announcement
OpenCode
View originalReddit
Step 5 Preview: Community Launch Overview
techspecsmart
View originalStep 5 Preview: Command Code Availability
maedahbatool
View originalStep 5 Preview: Architecture And Weight-Release Discussion
adefa
View originalYouTube
Step 5 Preview: Coding And Reasoning Tests
AICodeKing
View originalStep 5 Preview: Interactive Application Review
AI产品狙击手 kzhu
View originalStep 5 Preview: Agentic And Finance Demonstration
Code With Yousaf
View original