Migrate from OpenAI to StepFun - StepFun Documentation

Prerequisites

Stepfun models support both personal and enterprise use. Create an API key in the console and call with that key. For quick evaluation, you can also try models directly in the Experience Center.

Model selection

Stepfun offers multiple model families. Choose the counterpart that best matches what you use today.

Reasoning and multimodal models

step-3.7-flash: flagship multimodal reasoning model. Native image and video understanding, three reasoning effort levels (low/medium/high) — well-suited to agent, coding, and multimodal workloads.
step-3.5-flash: flagship language reasoning model. Top-tier reasoning quality with fast, reliable execution — suitable for logical reasoning, math, software engineering, and deep research.

Image generation

step-image-edit-2: unified text-to-image and image-editing model; a single edit completes in 1-2 seconds.

API compatibility

Stepfun models are compatible with the OpenAI API spec. You can call them directly via the OpenAI SDK. If your app already uses an OpenAI SDK, migration requires minimal changes. Supported OpenAI-compatible APIs:

SDK migration guide

OpenAI Python SDK

Replace api_key with your Stepfun API key and add base_url="https://api.stepfun.ai/v1". Set the model name to a Stepfun model.

copy

# Before
client = OpenAI(
    api_key="OPENAI_KEY"
)

# After
client = OpenAI(
    api_key="STEP_API_KEY",
    base_url="https://api.stepfun.ai/v1"
)

OpenAI TypeScript SDK

Replace apiKey with your Stepfun API key and add baseURL: "https://api.stepfun.ai/v1". Set the model name to a Stepfun model.

copy

// Before
const openai = new OpenAI({
	apiKey: 'OPENAI_KEY',
})

// After
const openai = new OpenAI({
	apiKey: 'STEP_API_KEY',
	baseURL: 'https://api.stepfun.ai/v1',
})

LangChain

When using LangChain, update openai_api_key, openai_api_base, and model_name in ChatOpenAI to switch to Stepfun.

copy

from langchain.chat_models import ChatOpenAI
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain
from os import getenv
from dotenv import load_dotenv

load_dotenv()

template = """You are the Stepfun assistant. Think step by step and answer based on the user's question: {question}"""

prompt = PromptTemplate(template=template, input_variables=["question"])

# Before
# llm = ChatOpenAI(
#   openai_api_key=getenv("OPENAI_KEY"),
#   model_name="gpt-4o"
# )

# After
llm = ChatOpenAI(
  openai_api_key=getenv("STEP_API_KEY"),
  openai_api_base="https://api.stepfun.ai/v1",
  model_name="step-3.7-flash"
)

llm_chain = LLMChain(prompt=prompt, llm=llm)

question = "How can Stepfun models help employees be more productive?"

print(llm_chain.run(question))

LangChain.js

For LangChain.js, update modelName, openAIApiKey, and basePath when initializing ChatOpenAI.

// Before
const chat = new ChatOpenAI({
	modelName: 'gpt-4o',
	openAIApiKey: $OPENAI_KEY,
})

// After

const chat = new ChatOpenAI(
	{
		modelName: 'step-3.7-flash',
		openAIApiKey: $STEP_API_KEY,
	},
	{
		basePath: 'https://api.stepfun.ai/v1',
	},
)

​Prerequisites

​Model selection

​Reasoning and multimodal models

​Image generation

​API compatibility

​SDK migration guide

​OpenAI Python SDK

​OpenAI TypeScript SDK

​LangChain

​LangChain.js

Prerequisites

Model selection

Reasoning and multimodal models

Image generation

API compatibility

SDK migration guide

OpenAI Python SDK

OpenAI TypeScript SDK

LangChain

LangChain.js