Muse Wonder

Name: Muse Wonder
Author: Meta

Multimodal

Meta

Muse Wonder is a proprietary reasoning model from Meta Superintelligence Labs, released on October 12, 2026 for personal and professional agents and long-running coding tasks. Active Memory retains relevant context across interactions, and Active Focus supports sustained task execution and error recovery. Muse Wonder accepts text, images, video, audio, and PDFs and produces text with a 1,048,576-token context window. It is available through the Meta Model API as muse-wonder-1.0 and in Muse Code.

Key Specifications

Parameters

Context

1.0M

Release Date

October 12, 2026

Average Score

52.8%

API Documentation Research Paper Results Blog

Timeline

Key dates in the model's history

Announcement

October 12, 2026

Last Update

October 13, 2026

Today

October 17, 2026

Technical Specifications

Parameters

Training Tokens

Knowledge Cutoff

Family

Capabilities

MultimodalZeroEval

Pricing & Availability

Input (per 1M tokens)

$4.00

Output (per 1M tokens)

$20.00

Max Input Tokens

1.0M

Max Output Tokens

Supported Features

Function CallingStructured OutputCode ExecutionWeb SearchBatch InferenceFine-tuning

Benchmark Results

Model performance metrics across various tests and benchmarks

Other Tests

Specialized benchmarks

Artificial Analysis

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Artificial Analysis Intelligence Index v4.3, independently evaluated by Artificial Analysis and reproduced in Meta's model card. Reported on a 0-100 scale, stored as a fraction. • Self-reported

52.0%

AutomationBench-AA

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Artificial Analysis strict pass@1; every deterministic end-state assertion must pass. Reported on a 0-100 scale, stored as a fraction. • Self-reported

39.1%

DeepSWE 1.1

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Datacurve third-party pass@1 results; 113 tasks across 91 repositories and five languages, graded with functional and regression tests. Reported on a 0-100 scale, stored as a fraction. • Self-reported

78.8%

FocusBench

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Internal personal agent evaluation across housing, travel, IT, and finance; persistence through blockers and adaptation across interactions and changing environments. Reported on a 0-100 scale, stored as a fraction. Score from the research blog chart; methodology: https://research.meta.ai/static/muse-wonder-model-card. • Self-reported

44.9%

GDP.pdf

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Professional tasks grounded in supplied PDF material; score reported by Surge. Reported on a 0-100 scale, stored as a fraction. • Self-reported

30.8%

Harvey LAB-AA

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Artificial Analysis result, version 1.1; share of legal deliverables meeting every criterion and passing hallucination checks. Reported on a 0-100 scale, stored as a fraction. • Self-reported

10.0%

HealthBench Professional

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. 525 professional health questions, no tools, GPT-5.4 rubric grader; length-normalized weighted rubric score in a common internal framework. Reported on a 0-100 scale, stored as a fraction. • Self-reported

60.5%

Job Bench

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. 65 professional tasks across 35 occupations; mean task-specific rubric reward. Reported on a 0-100 scale, stored as a fraction. • Self-reported

70.5%

MemoryBench

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Internal recall evaluation over 64K-100K-token conversations containing distractors, conflicting facts, and later updates; LLM judge with weighted question-specific rubrics. Reported on a 0-100 scale, stored as a fraction. • Self-reported

85.6%

OSWorld 2.1 (partial)

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. 108 desktop workflows in a common internal framework with official environments and graders; mean per-task partial reward. Reported on a 0-100 scale, stored as a fraction. • Self-reported

72.6%

OSWorld 2.1 (strict)

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. 108 desktop workflows in a common internal framework with official environments and graders; strict binary completion, separate from partial reward. Reported on a 0-100 scale, stored as a fraction. • Self-reported

34.3%

Personal Superintelligence Index

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Internal composite of Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT, and DRACO. Reported on a 0-100 scale, stored as a fraction. • Self-reported

54.3%

Program Bench

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. All models use mini-swe-agent; share of rollouts passing at least 95% of hidden behavioral tests; internet disabled after setup. Reported on a 0-100 scale, stored as a fraction. • Self-reported

50.0%

SWE-Marathon

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Abundant AI pass@1 results on long-running software projects. External results retain their reporting harnesses; not a controlled common-harness comparison. Reported on a 0-100 scale, stored as a fraction. • Self-reported

53.1%

Terminal-Bench 4.0-AA

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Artificial Analysis reported accuracy on terminal tasks with executable final-container-state verifiers. Reported on a 0-100 scale, stored as a fraction. • Self-reported

52.0%

Vals Finance Agent v2

Meta Muse Wonder model card, October 12, 2026; max reasoning effort. Vals AI private-test-set accuracy on financial analyst research tasks. Reported on a 0-100 scale, stored as a fraction. • Self-reported

56.6%

License & Metadata

License

proprietary

Announcement Date

October 12, 2026

Last Updated

October 13, 2026

Similar Models

All Models

Muse Wonder Full

Meta

Released:Oct 2026

Muse Spark 1.3

Meta

Released:Sep 2026

Price:$1.25/1M tokens

Muse Spark 1.2

Meta

Released:Aug 2026

Price:$1.25/1M tokens

Muse Spark 1.1

Meta

Released:Jul 2026

Price:$1.25/1M tokens

Muse Spark

Meta

Best score:0.9 (TAU)

Released:Apr 2026

Llama 4 Maverick

Meta

MM400.0B

Best score:0.9 (MMLU)

Released:Apr 2025

Price:$0.27/1M tokens

Llama 4 Scout

Meta

MM109.0B

Best score:0.8 (MMLU)

Released:Apr 2025

Price:$0.18/1M tokens

Muse Glimmer-30B

Meta

MM29.6B

Best score:0.8 (GPQA)

Released:Aug 2026

Recommendations are based on similarity of characteristics: developer organization, multimodality, parameter size, and benchmark performance. Choose a model to compare or go to the full catalog to browse all available AI models.