A one-stop large language model platform offering Text Generation, Deep Thinking, and Visual Understanding. Fully compatible with the OpenAI SDK — just change the base URL and API key. One API key unlocks AppBuilder, ModelBuilder, and AI Open Capabilities simultaneously.
ernie-4.5
deepseek-v4
GLM
Qwen
Kimi
System Prompt
Temperature
0.7
Max Tokens
2048
Capability
TextVisionFunc Call
Quick Start Guide
Four steps from zero to your first API call — no migration needed if you already use OpenAI SDK
01
Register & Get API Key
Sign up for a Baidu AI Cloud International account, then navigate to IAM → API Key Management to create a new key. Your key starts with bce-v3/ and must be prefixed with Bearer in requests. One key supports all three capabilities: AppBuilder, ModelBuilder, and AI Open Capabilities.
Qianfan is fully compatible with the OpenAI SDK. If you already use it, skip this step. Otherwise install with pip. No proprietary SDK needed — just the standard OpenAI Python package.
pip install openai
03
Configure & Call
Set base_url to api.baiduqianfan.ai/v1 and pass your API key. The appid header is optional. Choose any model from the pricing table below — ernie-4.5, deepseek-v4-pro, GLM, Qwen, and more.
04
Explore Capabilities
Beyond basic chat, try Structured Output for JSON responses, Function Calling for agent workflows, and Visual Understanding for image-based tasks. All through the same API endpoint — just change the model and parameters.
Platform Capabilities
Three core model capabilities plus advanced technical features
F01
Text Generation
Supports writing, dialogue, summarization, and Q&A. Use system and user roles in the messages array to control behavior. Example models: ernie-4.5, ernie-3.5-8k, ernie-4.0-turbo-8k.
F02
Deep Thinking
Advanced reasoning mode for complex problem solving — decision support, multi-step reasoning, and analytical tasks that require deeper cognitive processing.
F03
Visual Understanding
Object recognition, scene understanding, and image description. Process images alongside text for multimodal tasks including document analysis and visual Q&A.
F04
Structured Output
Enforce JSON schema responses for reliable, parseable output. Ideal for building data pipelines, API integrations, and automated workflows that require consistent formatting.
F05
Function Calling
Define tools and functions that the model can call autonomously. Build agents that interact with external APIs, databases, and services — with full parameter extraction and validation.
F06
Dialogue Prefix Continuation
Provide a partial assistant response and let the model continue from there. Useful for controlled generation, prompt chaining, and building conversational flows with predefined starting points.
Model Pricing & Discounts
Current promotional discounts across all supported models — up to 60% off with additional volume-based savings
Model
Discount
Max Discount (+10%)
Savings Range
ernie-4.5
40% OFF
50% OFF
40% – 50%
ernie-x1.1
10% OFF
20% OFF
10% – 20%
ernie-5.1
30% OFF
40% OFF
30% – 40%
deepseek-v3.2
40% OFF
50% OFF
40% – 50%
deepseek-v4-pro
60% OFF
70% OFF
60% – 70%
GLM
45% OFF
55% OFF
45% – 55%
Qwen
50% OFF
60% OFF
50% – 60%
Kimi
40% OFF
50% OFF
40% – 50%
GPT
25% OFF
35% OFF
25% – 35%
Note:Base discounts apply to all users. The additional +10% discount range is available for volume-based commitments and enterprise agreements. Discounts are promotional and subject to change. Contact sales for custom pricing.
Code Examples — Copy & Run
Fully compatible with OpenAI SDK. Just swap the base_url and api_key — your existing code works as-is.
Python — OpenAI SDK
from openai import OpenAI
client = OpenAI(
api_key="bce-v3/ALTAK-xxxxxxxx/xxxxxxxxxxxxxxxx", # Your Qianfan API Key
base_url="https://api.baiduqianfan.ai/v1", # Qianfan endpoint
default_headers={"appid": "app-xxxxxx"} # Optional: your appid
)
completion = client.chat.completions.create(
model="ernie-4.0-turbo-8k", # Or: ernie-4.5, deepseek-v4-pro, GLM, Qwen...
messages=[
{"role": "system", "content": "You are a helpful assistant."},
{"role": "user", "content": "Hello!"}
]
)
print(completion.choices[0].message)
Full API reference, model list, parameter descriptions, and integration guides. Follow this doc for the most up-to-date configuration methods and usage details.