outrun-ai (Python)

Python client for the Outrun API. Outrun is a prepaid, OpenAI-compatible API: one key for GPT, Claude, Gemini, Grok, DeepSeek, GLM, Kimi, Qwen and MiniMax models. This package is a thin wrapper over the official openai package with the base URL and API key set for you, so everything the OpenAI SDK does still works.

Install

pip install outrun-ai

Quickstart

Create a key at https://outrun.to/key, then:

export OUTRUN_API_KEY="your-key"

from outrun_ai import Outrun

client = Outrun()  # reads OUTRUN_API_KEY
print(client.ask("gpt-5.6-terra", "Say hello in five words."))
print(client.ask("deepseek-v4.1-flash", "Name three prime numbers.", system="Be brief."))

You can also pass Outrun(api_key="...", base_url="..."). The default base URL is https://api.outrun.to/v1. If no key is found, the client raises OutrunError.

Streaming

for text in client.ask_stream("deepseek-v4.1-flash", "Write a haiku about rain."):
    print(text, end="", flush=True)

Model ids and prices

prices() reads the public price list. It needs no key and costs nothing.

from outrun_ai import prices

for m in prices(kind="text"):
    tier = m["tiers"][0]
    print(m["id"], tier["prices"].get("input"), tier["prices"].get("output"), "USD per", tier["unit"])

Each item has id (the model name to send), name, family, kind (text, image, video or audio), unit, tiers and paused. Paused models are left out unless you pass include_paused=True. The same list is at https://outrun.to/pricing.

Everything else

Outrun subclasses openai.OpenAI, so the usual calls work unchanged:

client.chat.completions.create(
    model="deepseek-v4.1-flash",
    messages=[{"role": "user", "content": "Hi"}],
)
client.models.list()

AsyncOutrun is the async version (await client.ask(...), async for t in client.ask_stream(...)).

Each API request carries an X-Outrun-SDK: python/<version> header. Nothing else is added.

Links

• Docs: https://outrun.to/docs/
• LangChain: pip install langchain-outrun (source)
• Support: support@outrun.to

License: MIT.