cai/docs/providers/minimax.md

MiniMax Configuration

MiniMax AI

MiniMax provides large language models with an OpenAI-compatible API. CAI includes a built-in MiniMaxProvider for easy integration.

Setup

1. Get an API key from MiniMax.
2. Set the environment variable:

export MINIMAX_API_KEY=<your-api-key>

Usage

Option 1: Using MiniMaxProvider (Recommended)

from cai.sdk.agents import Agent, Runner, RunConfig
from cai.sdk.agents.models.minimax_provider import MiniMaxProvider

provider = MiniMaxProvider()

agent = Agent(
    name="assistant",
    instructions="You are a helpful assistant.",
)

result = await Runner.run(
    agent,
    "Hello!",
    run_config=RunConfig(model_provider=provider),
)

Option 2: Using environment variables with LiteLLM

CAI_MODEL=openai/MiniMax-M2.7
MINIMAX_API_KEY=<your-api-key>
OPENAI_API_BASE=https://api.minimax.io/v1

Option 3: Direct model on Agent

from openai import AsyncOpenAI
from cai.sdk.agents import Agent, OpenAIChatCompletionsModel

client = AsyncOpenAI(
    api_key="<your-api-key>",
    base_url="https://api.minimax.io/v1",
)

agent = Agent(
    name="assistant",
    instructions="You are a helpful assistant.",
    model=OpenAIChatCompletionsModel(
        model="MiniMax-M2.7",
        openai_client=client,
    ),
)

Available Models

Model	Context Window	Description
MiniMax-M2.7	1M tokens	Latest and most capable model
MiniMax-M2.5	1M tokens	Strong general-purpose model
MiniMax-M2.5-highspeed	204K tokens	Optimized for speed

Notes

• MiniMax uses an OpenAI-compatible API, so it works with OpenAIChatCompletionsModel.
• Tracing is sent to OpenAI by default. If you don't have an OpenAI API key, disable tracing with set_tracing_disabled(True) or set a tracing-specific key.
• Temperature range: [0.0, 1.0].