Overview
StepFun is an AI model platform from StepFun (阶跃星辰) for developers and teams building chat, coding, agent, speech, vision, and image-generation workflows. Its global platform combines a browser-based AI Studio with an API catalog that uses one account and key across several model families.
The current flagship is Step 5 Preview, a sparse mixture-of-experts model designed for long-horizon agentic work, software engineering, finance, and professional knowledge tasks. It supports a 1-million-token context window, up to 64K output tokens, and text, image, and video input. The broader StepFun catalog also includes faster language models plus speech and image services, so teams can use the platform as a multi-model layer rather than a single chatbot.
StepFun offers two distinct paid access paths. Standard API usage is billed by tokens or media units. Step Plan is a separate credit-based subscription for coding clients such as Claude Code, OpenClaw, and Cursor. Teams should choose the billing channel that matches their workflow instead of assuming Step Plan credits can be spent through the standard API.
Key Features
Unified model API - Use one platform and API key for language, reasoning, vision, speech, and image-generation models.
Step 5 Preview flagship - Build long-context agents with a 1M-token window, up to 64K generated tokens, and text, image, or video inputs.
Agent integration features - Use streaming responses, tool calling, JSON Mode, JSON Schema, prompt caching, and adjustable reasoning effort. External search and code execution still need to be supplied by the integrating application.
AI Studio - Test supported models in a browser before wiring them into an application or agent framework.
OpenAI-compatible endpoint - Connect compatible SDKs and tools through the documented
https://api.stepfun.ai/v1base URL.Step Plan for coding clients - Use a monthly credit plan with supported developer clients when subscription-style access fits better than metered API billing.
How to Get Started
Create an account on the StepFun platform, open AI Studio to test the model you need, then generate an API key for programmatic access. The official examples use an OpenAI-compatible base URL, which reduces migration work for applications already built around that request format.
For an agent prototype, start with a constrained tool set and Step 5 Preview's medium reasoning-effort setting for general multi-step agent tasks, using low for simple tasks and high for complex planning. StepFun's model can emit tool calls, but the surrounding application is responsible for providing, authorizing, and executing tools such as web search or code execution.
Developers using Claude Code, Cursor, OpenClaw, or another supported coding client can also evaluate Step Plan. Keep its Base URL and usage accounting separate from the standard API channel, as the two access paths use different quotas and billing systems.
Pricing & Plans
StepFun publishes pay-as-you-go API pricing and separate Step Plan subscriptions.
Standard API
For Step 5 Preview, the official global pricing page lists:
| Usage | Price per 1M tokens |
|---|---|
| Cache-miss input | $1.00 |
| Cache-hit input | $0.05 |
| Output | $2.70 |
Reasoning tokens and final-answer tokens are both billed as output. StepFun also publishes usage tiers: the entry V0 tier has 5 concurrent requests, 10 requests per minute, and 5 million tokens per minute, while higher top-up tiers raise those limits.
Step Plan
| Plan | Monthly price | Included credits |
|---|---|---|
| Flash Mini | $6.99 | 400M |
| Flash Plus | $9.99 | 1,600M |
| Flash Pro | $29 | 8,000M |
| Flash Max | $99 | 40,000M |
Step Plan credits are designed for supported coding clients and are not interchangeable with pay-as-you-go API usage. Prices, model availability, and limits can change, so confirm the active documentation before committing a production workload.
Best For
- Developers building AI agents that need long context, tool calling, structured outputs, and multimodal evidence
- Engineering teams evaluating an AI code generator for repository-scale work and terminal-based workflows
- Product teams that want one API for chat, vision, speech, and image-generation features
- Researchers and analysts working with long reports, financial material, images, or video
- Coding-client users who prefer a monthly Step Plan over purely metered API billing
FAQ
What is StepFun?
StepFun is an AI model company and platform. Its global developer service provides language, vision, speech, and image models through AI Studio and an API, while Step Plan offers a separate subscription channel for supported coding clients.
What is Step 5 Preview?
Step 5 Preview is StepFun's current flagship model for agentic work, software engineering, finance, and professional knowledge tasks. It supports a 1M-token context window, up to 64K output tokens, and text, image, and video inputs.
Is Step 5 Preview open source?
The weights are not open as of September 22, 2026. StepFun says Step 5 Preview will be released with open weights on October 15, 2026; this does not by itself establish an open-source license. Until an official repository is available, users should treat the hosted platform and API as the verified access paths.
Does StepFun include web search and code execution?
The API supports tool calling, but the model documentation says search and code execution are provided by the integrating application. Developers need to connect and govern those tools themselves.
Sources
- StepFun global platform - model catalog, API positioning, and AI Studio access.
- Step 5 Preview model documentation - context, output, modalities, supported features, and media limits.
- StepFun API pricing - token prices and usage tiers.
- Step Plan overview - subscription tiers, coding-client support, and billing-channel separation.