M

Meta Model API

Meta's model inference API for developers, offering direct, self-serve, and affordable access to models such as Muse Wonder and Muse Spark.

paid
Last Updated 11/2/2026
Visit Website

About Meta Model API

Meta Model API is Meta's inference API, providing direct access to Meta's latest models such as Muse Wonder and Muse Spark. Developers can call the API to build their own apps or agents, or use Muse Code to get a ready-made coding agent in the terminal—both use the same models and authentication. Muse Wonder is what Meta officially calls its strongest model, built for complex software engineering and long-running agent work: it can make and verify changes across multiple files in real codebases in parallel, delegate multi-step tasks to sub-agents, and handle long-running tasks through parallel tool calls. The API is billed by token, by image, by audio duration, and similar usage-based methods, with no minimum spend or upfront commitment.

Best for

Developers who need to call Meta Muse models in their own apps or agents
Engineers who need a multi-agent coding tool in the terminal
Teams that need to handle complex engineering tasks in large codebases

Best use cases

Call Muse models via the API in your own apps or agents

Use Muse Code in the terminal for multi-agent coding

Make and verify code changes in parallel across multiple files in large codebases

Let agents operate browsers, desktop, and mobile apps to complete everyday tasks

Use Muse models through OpenRouter in existing tools and workflows

Features

Complex engineering tasks

Solve complex engineering tasks in large codebases, making and verifying changes across multiple files in parallel.

Long-running agents

Hand off multi-step tasks that run for hours without interruption; Muse Wonder plans the work, delegates parts of it to sub-agents, calls tools in parallel, and recovers from errors to keep critical work moving forward.

Advanced computer use

Power agents that can view and operate browsers, desktop, and mobile apps to complete everyday tasks.

Compatible with mainstream API formats

Provides OpenAI-compatible Responses and Chat Completions endpoints, as well as an Anthropic Messages-compatible endpoint, making it easy to plug into existing toolchains.

Pros

  • Provides OpenAI-compatible and Anthropic Messages-compatible endpoints, making it easy to integrate with existing tools and code
  • Supports long-running agents that can plan tasks, delegate to sub-agents, call tools in parallel, and recover from errors
  • Usage-based billing with no minimum spend or upfront commitment
  • Beyond text models, also offers image generation, speech transcription, and image/video segmentation capabilities

FAQ

Model API is billed by usage: text models are billed by token, image generation by image, Muse Voice Transcribe by the number of audio hours processed, and Segment Anything Model (SAM) by the images or video frames segmented, with no minimum spend or upfront commitment.
Rate limits are calculated per team rather than per API key: requests, tokens, and streams from all keys within a team count toward the same limit, and each model has its own independent limit, so traffic to one model does not consume another model's quota.
The API is authenticated via a Bearer token (MODEL_API_KEY); keys can be created and managed in the Model API console.

Tool Information

DeveloperMeta Model API
Release Date11/2/2026
Pricingpaid
Visit Website