# Drop-in replacement for OpenAI. Change two lines.from openai import OpenAI
client = OpenAI(
base_url="https://api.corriente.ai/v1",
api_key="ck_your_key_here"
)
response = client.chat.completions.create(
model="auto", # Smart routing — or pick any of 779 models
messages=[{"role": "user", "content": "Hello, Corriente"}]
)
print(response.choices[0].message.content)
Built different.
Smart Routing
Set model to "auto" and our Polymorphic Brain picks the best model for each query. Code goes to coding models. Math goes to math models. Reasoning goes to reasoning models. Automatically.
OpenAI Compatible
Change your base_url and api_key. That's it. Works with every OpenAI SDK, LangChain, LlamaIndex, and any tool that speaks the OpenAI API format.
Bare Metal
7 NVIDIA DGX nodes. 128GB unified memory each. No shared cloud instances. No noisy neighbors. Your inference runs on dedicated hardware.
Your Data Stays Here
We don't log prompts. We don't train on your data. We don't sell your queries. Your conversations are yours. Period.
Function Calling
Full tool use and function calling support. Build agents that take actions, call APIs, and interact with external systems.
Conversation Memory
Built-in session management. Create a session, send messages, and the API remembers context across turns. No client-side history management needed.
779 models. One key.
From 0.5B fast models to 236B reasoning powerhouses. We've got the model for the job.
pip install openai
from openai import OpenAI
client = OpenAI(base_url="https://api.corriente.ai/v1", api_key="ck_your_key")
# Auto-route to best model
r = client.chat.completions.create(model="auto", messages=[{"role":"user", "content":"Hello"}])
# Or pick a specific model
r = client.chat.completions.create(model="qwen2.5-coder:32b", messages=[...])
# Use sessions for multi-turn# POST /v1/sessions to create, then pass X-Session-ID header