SDKs and libraries

Build with the clients you already run. Meta Model API doesn't need a client of its own: use the OpenAI SDK for the Responses API and Chat Completions API, the Anthropic SDK for the Messages API, or any OpenAI-compatible client, such as LangChain, LlamaIndex, the Vercel AI SDK, or LiteLLM. Set the client's base URL, pass your Model API key, and keep the rest of your code.

Supported features

Python

shell

pip install openai

Python (OpenAI SDK)

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MODEL_API_KEY"],
    base_url="https://api.meta.ai/v1",
)

response = client.responses.create(
    model="muse-spark-1.3",
    input="Explain mixture-of-experts in two sentences.",
)
print(response.output_text)

TypeScript

shell

npm install openai

TypeScript (OpenAI SDK)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.MODEL_API_KEY,
  baseURL: "https://api.meta.ai/v1",
});

const response = await client.responses.create({
  model: "muse-spark-1.3",
  input: "Explain mixture-of-experts in two sentences.",
});
console.log(response.output_text);

Anthropic SDK

Use the Anthropic SDK for the Model API Messages API. Set base_url to https://api.meta.ai (the SDK appends /v1/messages), and pass your Model API key as auth_token, which the SDK sends as a bearer token.

A response's content starts with a redacted_thinking block that carries the model's encrypted reasoning, so read the text blocks instead of indexing content[0]:

Python (Anthropic SDK)

import os
from anthropic import Anthropic

client = Anthropic(
    base_url="https://api.meta.ai",
    auth_token=os.environ["MODEL_API_KEY"],
)

message = client.messages.create(
    model="muse-spark-1.3",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Explain mixture-of-experts in two sentences."}],
)
print("".join(block.text for block in message.content if block.type == "text"))

OpenAI-compatible clients

Any client or framework that speaks the OpenAI API and accepts a custom base URL works with Model API, including LangChain, LlamaIndex, the Vercel AI SDK, and LiteLLM. Most ask for three values: the base URL https://api.meta.ai/v1, your Model API key, and a model ID such as muse-spark-1.3.

LiteLLM reaches Model API through its OpenAI provider, so prefix the model ID with openai/:

shell

pip install litellm

Python (LiteLLM)

import os
import litellm

response = litellm.completion(
    model="openai/muse-spark-1.3",
    api_base="https://api.meta.ai/v1",
    api_key=os.environ["MODEL_API_KEY"],
    messages=[{"role": "user", "content": "Explain mixture-of-experts in two sentences."}],
)
print(response.choices[0].message.content)

Segment Anything Model client libraries

Segment Anything Model 3.1 (SAM 3.1) returns its boxes and masks as a token grammar in the Responses output text. The Meta-SAM client libraries decode it for you. @meta-sam/parser (TypeScript) and meta_sam_parser (Python) turn the stream into typed segmentation snapshots and decode masks. The optional @meta-sam/graphics, @meta-sam/video, and @meta-sam/react packages draw overlays and play segmented video.

Authentication

None of these clients reads MODEL_API_KEY on its own. The OpenAI SDK looks for OPENAI_API_KEY and the Anthropic SDK looks for ANTHROPIC_API_KEY, so pass your Model API key explicitly when you create the client:

Python (OpenAI SDK)

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["MODEL_API_KEY"],
    base_url="https://api.meta.ai/v1",
)

Supported features#

Python#

TypeScript#

Anthropic SDK#

OpenAI-compatible clients#

Segment Anything Model client libraries#

Authentication#

Try it live

Supported features

Python

TypeScript

Anthropic SDK

OpenAI-compatible clients

Segment Anything Model client libraries

Authentication