Skip to content

Muse

Muse is a family of large language models (LLMs) and AI products from Meta Superintelligence Labs that spans the hosted Muse Wonder and Muse Spark models and the open-weights Muse Glimmer and Muse Spark Open.

Meta introduced Muse Spark in April 2026, and version 1.3 from September 2026 is the latest Spark release. Spark is a multimodal reasoning model that reads text, images, video, and PDFs within a context window of about one million tokens. Reasoning can’t be switched off, but an effort setting trades depth for speed and cost.

In October 2026, Meta added Muse Wonder, a larger frontier model served through the Meta Model API, and Muse Wonder Full, which is available only to paid Muse Code subscribers. Neither Wonder model has downloadable weights.

Later that month, Meta released Muse Spark Open, an open-weights version of Spark with 816 billion total parameters, of which 42 billion are active per token. It ships under the Muse Model License, which has limited commercial terms, and runs on third-party hosts rather than the Meta Model API.

Python developers call Spark through the Meta Model API, which is OpenAI-compatible, so the openai package works once it points at Meta’s base URL. The API supports streaming, image input, tool calling, and structured output. Spark is also available through OpenRouter at the same token prices.

Billing is pay-as-you-go per token. A cheaper contributor tier, selected by a model ID ending in -contributor, lets Meta train on the prompts and completions sent to it.

Muse Glimmer is a 30-billion-parameter model distilled from Spark and released in August 2026 under the Apache 2.0 license. Its weights download from Hugging Face and run locally under Ollama, vLLM, or llama.cpp, since Glimmer isn’t served on the Meta Model API.

The family also includes Muse Code, a terminal coding agent, along with Muse Image and Muse Voice Transcribe. On its own, the Muse name also refers to Meta’s consumer personal agent, launched in the US in September 2026. It runs on Spark but has no public API.

Official website: dev.meta.ai

How to Use the OpenRouter API to Access Multiple AI Models via Python

Tutorial

How to Use the OpenRouter API to Access Multiple AI Models via Python

Access models from popular AI providers in Python through OpenRouter's unified API with smart routing, fallbacks, and cost controls.

intermediate ai api

For additional information on related topics, take a look at the following resources:

Have a question about this? Mentor AI can show you examples, compare related terms, and point you to tutorials.


By Martin Breuss • Updated Nov. 5, 2026