openai-oxide for Python

Python bindings for openai-oxide — the fastest OpenAI client. Also available on crates.io (Rust) and npm (Node.js).

Rust core compiled to native Python extension via PyO3 + maturin.

Install

pip install openai-oxide
# or
uv pip install openai-oxide
# or
uv add openai-oxide

For local development (requires Rust toolchain):

cd openai-oxide-python
uv sync
uv run maturin develop --release

Usage

import asyncio, json
from openai_oxide import Client

async def main():
    client = Client()  # reads OPENAI_API_KEY env var

    # Simple
    r = await client.create("gpt-4o-mini", "Hello!")
    print(r["text"])

    # Streaming
    stream = await client.create_stream("gpt-4o-mini", "Explain quantum computing...", max_output_tokens=200)
    async for event in stream:
        print(event)

    # Structured output
    schema = json.dumps({
        "type": "object",
        "properties": {"answer": {"type": "integer"}},
        "required": ["answer"],
        "additionalProperties": False
    })
    r = await client.create_structured("gpt-4o-mini", "What is 7*8?", "math", schema)
    print(json.loads(r["text"])["answer"])  # 56

    # Function calling
    tools = json.dumps([{
        "name": "get_weather",
        "description": "Get weather",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
            "additionalProperties": False
        }
    }])
    r = await client.create_with_tools("gpt-4o-mini", "Weather in Tokyo?", tools)
    print(r["function_calls"])

asyncio.run(main())

Test

uv run python test_oxide.py

Benchmarks

Run uv run python examples/bench_python.py to compare against the official openai SDK.

See the main README for full benchmark results.