Muse Glimmer-30B

Name: Muse Glimmer-30B
Author: Meta

Multimodal

Meta

Muse Glimmer-30B is Meta Superintelligence Labs' open-weight, dense multimodal model for autonomous agentic work on consumer hardware.

Key Specifications

Parameters

29.6B

Context

Release Date

August 10, 2026

Average Score

65.1%

API Documentation Model Weights Results Blog

Timeline

Key dates in the model's history

Announcement

August 10, 2026

Last Update

August 29, 2026

Technical Specifications

Parameters

29.6B

Training Tokens

Knowledge Cutoff

January 4, 2026

Family

Capabilities

MultimodalZeroEval

Benchmark Results

Model performance metrics across various tests and benchmarks

Programming

Programming skills tests

SWE-Bench Verified

High reasoning. • Self-reported

76.0%

Reasoning

Logical reasoning and analysis

GPQA

High reasoning; Diamond split; AA evaluation. • Self-reported

83.5%

Other Tests

Specialized benchmarks

AA-LCR

High reasoning. • Self-reported

80.0%

AIME 2026

High reasoning. • Self-reported

94.7%

Beam 128K

High reasoning; 128K context. • Self-reported

65.1%

CharXiv-R

High reasoning; CharXiv reasoning split. • Self-reported

78.8%

CI Memories Coverage

High reasoning; coverage. • Self-reported

64.8%

CI Memories Violation Rate

High reasoning; violation rate (lower is better). • Self-reported

73.6%

DeepSearchQA

High reasoning. • Self-reported

74.6%

GAIA2

High reasoning. • Self-reported

43.3%

Humanity's Last Exam

High reasoning; text only; no tools. • Self-reported

22.0%

IFBench

High reasoning. • Self-reported

77.0%

MCP Atlas

High reasoning; public set. • Self-reported

75.5%

MMMU-Pro

High reasoning. • Self-reported

74.0%

OmniDocBench 1.5

High reasoning. • Self-reported

75.8%

OSWorld-Verified

High reasoning. • Self-reported

65.9%

SciCode

High reasoning. • Self-reported

43.6%

ScreenSpot Pro

High reasoning. • Self-reported

75.4%

Siren AgentDojo Attack Success Rate

High reasoning; attack success rate (lower is better). • Self-reported

71.6%

Siren AgentDojo Utility

High reasoning; utility. • Self-reported

94.2%

SkillsBench

High reasoning; with skills. • Self-reported

44.3%

SWE-Bench Pro

High reasoning. • Self-reported

51.2%

Tau3 Banking

High reasoning. • Self-reported

23.5%

Terminal-Bench 2.1

High reasoning; Terminus-2 harness. • Self-reported

51.7%

WildClawBench

High reasoning. • Self-reported

47.6%

License & Metadata

License

apache-2.0

Announcement Date

August 10, 2026

Last Updated

August 29, 2026

Compare Muse Glimmer-30B

All comparisons

vs Muse Spark 1.2 vs Muse Spark 1.1 vs Nemotron 3 Ultra (550B A55B)vs GPT-5.6 Sol vs Qwen3.8-27B vs Seed 2.1 Pro vs Grok 4.6 vs Seed 2.1 Turbo

Similar Models

All Models

Llama 3.2 90B Instruct

Meta

MM90.0B

Best score:0.9 (MMLU)

Released:Sep 2024

Price:$1.20/1M tokens

Llama 3.2 11B Instruct

Meta

MM10.6B

Best score:0.7 (MMLU)

Released:Sep 2024

Price:$0.18/1M 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

Qwen3.6-35B-A3B

Alibaba

MM35.0B

Best score:0.9 (GPQA)

Released:Apr 2026

Qwen3.8-27B

Alibaba

MM27.8B

Best score:0.9 (GPQA)

Released:Aug 2026

Muse Spark 1.2

Meta

Released:Aug 2026

Price:$0.10/1M tokens

Muse Spark 1.1

Meta

Released:Jul 2026

Price:$1.25/1M tokens

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.