Muse Spark OpenOpen Weight

Released

Muse Spark Open is an AI model released by Meta on Thursday, Oct 15 2026, 3 days after Muse Wonder Full. It is an open-weight model — the trained weights are available to download and run. It is a 816B parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model with a 1M token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover MCP Atlas, BabyVision (with tools), HLE-Diamond (no tools), DeepSWE 1.1, Next.js Evals, Terminal-Bench 2.1, and 4 more.

Available from

BaseTen$1.25$4.251Mfp4
Makora$1.25$4.251Mfp4
Modal$1.25$4.251M—
Nebius$1.25$4.251Mfp4
Together$1.25$4.251M—
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked November 9, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

Coding

DeepSWE 1.1Agentic coding — Artificial Analysis' independent test of deep, agentic software-engineering work — the AI has to plan and carry out substantial coding tasks end to end. (Version 1.1 of the test.) Higher is better.
60%
#28 of 42
Next.js EvalsNext.js coding — Vercel's open eval of how well AI coding agents build and migrate real Next.js apps — measured as the share of tasks the agent completes successfully. Higher is better.
93%
#8 of 35

Terminal & CLI

Terminal-Bench 2.1Agentic terminal coding — Can the AI work in a command-line terminal — running commands and finishing technical setup tasks the way a developer would? Higher is better.
76.4%
#24 of 39

Agentic & tool use

MCP AtlasMulti-step tool use — Can the AI chain together many tools and steps to complete one bigger task, rather than doing just a single thing? Higher is better.
89.2%
#1
JobBenchProfessional tool use — Tests the AI on professional workplace tasks that require using real work tools — the kind of multi-step jobs an office worker handles. Higher is better.
61.8%
#5 of 12

Reasoning & science

HLE-Diamond (no tools)Reasoning — accuracy on the 1,000-question HLE-Diamond reasoning and knowledge subset without tools
26.1%

Knowledge work

GDPval-AA v2Knowledge work — economically valuable knowledge work (v2, re-based Elo)
1401
#19 of 25

Multimodal

BabyVision (with tools)Visual reasoning — mean accuracy on 388 visual reasoning questions with tools
80.2%
ZeroBench (tools, pass@5)Visual reasoning — percentage of 100 visual problems solved at least once in five attempts with tools
50%
#2 of 3
SimpleVQA (with tools)Visual reasoning — accuracy on 2,024 English and Chinese visual questions with tools
73.6%
#2 of 3

Compare Muse Spark Open with

Muse Spark Open

About

Muse Spark Open was released on October 15, 2026 as Meta's downloadable multimodal model for coding and agent tasks. It had 816 billion total parameters, about 42 billion active per token and a one-million-token context window. The weights were distributed under the Muse Model License with limited commercial terms.

Meta demonstrated running it on two Apple M5 Ultra Mac Studios using MLX and also announced hosted inference partners. That hardware example was a specific deployment configuration, not a claim that the model would fit on any desktop. Developers could customize the weights, subject to the model's license and the infrastructure needed to serve it.

Frequently asked questions

Muse Spark Open was released by Meta on Thursday, Oct 15 2026.

All Meta releases

19 tracked