Benchmark Almanac

Muse Glimmer

Muse Glimmer ranks first on AA-LCR (Artificial Analysis Long Context Reasoning).

Meta Released August 10th, 2026 Compare with others
48
Benchmarks with a score
1
Ranked first, out of five or more models
1
In the top three
Meta Aug 2026

Muse Glimmer

Open weights

147.0 #116 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • MMMU-Pro 74% #44 /90
  • IFBench 77% #15 /80
  • SciCode 43.6% #23 /49

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
AA-LCR (Artificial Analysis Long Context Reasoning) 80% #1 of 38 Lab model card
IFBench 77% #15 of 80 Lab model card
LMArena (Elo) 1,425 #99 of 335 Official leaderboard
SEA-HELM 72.1 #27 of 71 Official leaderboard
MMLU-ProX 80.1% #11 of 23 Lab model card
Humanity's Last Exam 22% #100 of 154 Lab model card
Hemingway-bench 913 #41 of 42 Official leaderboard
Belebele 87.8% #1 of 1 Lab model card
BOUQuET 36.5 #1 of 1 Lab model card
FLORES+ 47.5 #1 of 1 Lab model card
Linguini 8.80% #1 of 1 Lab model card
Omnilingual-Gaia2 60.4% #1 of 1 Lab model card

Math and science

BenchmarkScoreRankSource
GPQA Diamond 83.5% #93 of 302 Lab model card
AIME 2026 94.7% #14 of 38 Lab model card
DTBench 85.6% #70 of 180 Official leaderboard
ARC-AGI 66.9% #15 of 21 Lab model card

Coding

BenchmarkScoreRankSource
SWE-bench Verified 76% #42 of 155 Lab model card
SciCode 43.6% #23 of 49 Lab model card
Code Arena 1,355 #102 of 128 Official leaderboard
SWE-Bench Pro 51.2% #67 of 81 Lab model card

Agents and tools

BenchmarkScoreRankSource
PinchBench 90.5% #4 of 14 Lab model card
ScreenSpot-Pro 75.4% #10 of 32 Lab model card
ScreenSpot-v2 93.3% #5 of 14 Lab model card
LLM Chess 214 #52 of 123 Official leaderboard
MCP Atlas 75.5% #23 of 46 Lab model card
τ³-bench Banking 23.5% #12 of 22 Lab model card
WildClawBench 47.6% #17 of 31 Official leaderboard
OSWorld-Verified 65.9% #25 of 41 Lab model card
SkillsBench 44.3% #12 of 14 Lab model card
DeepSearchQA 74.6% #20 of 23 Lab model card
Terminal-Bench 2.1 51.7% #50 of 56 Lab model card
WitnessBench 0.21% #11 of 12 Official leaderboard
Meta Agents Research Environments (ARE) / Gaia2 43.3% #3 of 3 Lab model card

Professional work

BenchmarkScoreRankSource
FoodTruck Bench $10,425 #39 of 59 Official leaderboard
Tuesday Work Index 27.6 #41 of 44 Official leaderboard
GDPval-AA v2 953 #26 of 27 Lab model card
DAYJOB: Finance 0% #39 of 39 Official leaderboard
DAYJOB: Healthcare 0% #37 of 37 Official leaderboard

Vision, audio and robotics

BenchmarkScoreRankSource
MMMU-Pro 74% #44 of 90 Lab model card
CharXiv 78.8% #39 of 75 Lab model card
OmniDocBench v1.5 75.8 #5 of 9 Lab model card
Chartography 17.2% #27 of 43 Official leaderboard
4DCodeBench 0.44 #12 of 16 Official leaderboard

Safety and security

BenchmarkScoreRankSource
CIMemories 26.4% #1 of 2 Lab model card
AgentDojo 28.4% #7 of 8 Lab model card
PrivacyLens 49.2% #1 of 1 Lab model card

Other

BenchmarkScoreRankSource
ACCoRD 72.0 #38 of 152 Official leaderboard
LMCA 34.9% #79 of 147 Official leaderboard

Its lab's numbers put it at 147.0, number 116 of 404

The Self-Reported Index fits one score to the 21 capability results Meta reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 145.2 to 148.9. How we compute it.

Its lab's number next to a leaderboard

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
WildClawBench 47.6% WildClawBench 47.6% Leaderboard, OpenClaw Same

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device (Meta) and model card Aug 10, 2026

Its system card

Every result in the longer documents the lab published with it. 13 of them aren't in the launch post.

Muse Glimmer 30B: An On-Device Agentic Model Oct 19, 2026

Capabilities

Safety

Robustness