mlx-optiq

Use with frameworks

OptiQ Cloud speaks OpenAI's chat completions format, so any library with an OpenAI-compatible adapter can call it. Give it the base URL and a key; there is nothing of ours to install.

SettingValue
Base URLhttps://cloud.mlx-optiq.com/api/v1
Keysk-optiq-cloud-… from the dashboard, here read from OPTIQ_CLOUD_API_KEY
Modelsglm-5.3-flash, deepseek-v4.1-flash, minimax-m3.1-flash, glm-5.3

Prices and context windows are on the overview, and GET /models returns them without a key. Fields, streaming and errors are in the API reference.

LiteLLM

The openai/ prefix picks LiteLLM's OpenAI-compatible adapter, which strips the prefix and sends the rest as the model id.

from litellm import completion
import os

reply = completion(
    model="openai/glm-5.3-flash",
    api_base="https://cloud.mlx-optiq.com/api/v1",
    api_key=os.environ["OPTIQ_CLOUD_API_KEY"],
    messages=[{"role": "user", "content": "Hello"}],
    reasoning_effort="low",
)

For the LiteLLM proxy, add the models to its config:

model_list:
  - model_name: glm-5.3-flash
    litellm_params:
      model: openai/glm-5.3-flash
      api_base: https://cloud.mlx-optiq.com/api/v1
      api_key: os.environ/OPTIQ_CLOUD_API_KEY

LangChain

from langchain_openai import ChatOpenAI
import os

llm = ChatOpenAI(
    model="deepseek-v4.1-flash",
    base_url="https://cloud.mlx-optiq.com/api/v1",
    api_key=os.environ["OPTIQ_CLOUD_API_KEY"],
)
print(llm.invoke("Hello").content)

LlamaIndex

Use OpenAILike from llama-index-llms-openai-like, and tell it the model is a chat model that can call tools.

from llama_index.llms.openai_like import OpenAILike
import os

llm = OpenAILike(
    model="glm-5.3-flash",
    api_base="https://cloud.mlx-optiq.com/api/v1",
    api_key=os.environ["OPTIQ_CLOUD_API_KEY"],
    is_chat_model=True,
    is_function_calling_model=True,
    context_window=1048576,
)

Vercel AI SDK

import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { generateText } from "ai";

const optiq = createOpenAICompatible({
  name: "optiq-cloud",
  baseURL: "https://cloud.mlx-optiq.com/api/v1",
  apiKey: process.env.OPTIQ_CLOUD_API_KEY,
});

const { text } = await generateText({ model: optiq("glm-5.3-flash"), prompt: "Hello" });

Reasoning

Every model thinks by default, and thinking is billed as output. Pass reasoning_effort as above to think less. GLM 5.3 Flash, GLM 5.3 and MiniMax M3.1 Flash cannot turn thinking off, so none is refused on them with a 400; DeepSeek V4.1 Flash takes it. The values each model accepts are under Reasoning.

Coding agents such as Claude Code, Codex and OpenCode are covered in Use with coding agents.