[whatmodel.sh](../../README.md) / Muse Wonder Full # Muse Wonder Full Meta · released 2026-10 · closed weights **score 81.7** ±1.4 · #7 of 732 · from 6 benchmark results across 2 sources Score within the [LMArena (text)](https://whatmodel.sh/?set=lmarena-text&score=set) comparison set only. ## What it's good at ● marks Muse Wonder Full among every model on that task or test, worst to best. 
task         vs the others        rank  counts  tests
agent tasks  ▁▁▃▆█▄▅▅▄▄●▁    #9 of 313     58%      3
coding       ▁▁▁▃▅▇█▇▅▄▃●   #7 of 469*     34%      1
agent tasks · #9 of 313 · 58% of its score Trails [Gemini 4 Argon](gemini-4-argon.md), [Claude Sonnet 5.5](claude-sonnet-5-5.md) here, though they rank lower overall. 
test                                    result  vs the others in its run  difficulty      counts  why
Arena Agent (October 2026 aggrega…  0.1009 IPS  ▁▁▁▃▃▂█▁●▂▁▁ beats 51/59  avg -0.379 IPS     49%  sharp
JobBench (maker)                         75.3%  █·█····█···● beats 3/3    hard · avg 25%    5.5%  at its level
OSWorld 2.1 (partial score) (make…       78.4%  ██·····●··██ beats 2/4    hard · avg 28%    3.5%  small run
coding · #7 of 469 · 34% of its score Beats [GPT-6.1 Terra](gpt-6-1-terra.md) here, though they rank higher overall. Trails [Claude Sonnet 5.5](claude-sonnet-5-5.md) here, though they rank lower overall. 
test                       result  vs the others in its run    difficulty  counts  why
Arena WebDev (Code Arena)  1725.3  ▁▁▃▂▅█▃▆▅▅●▂ beats 123/129  avg 981        34%  sharp
other tests (2) 
test                              result  vs the others in its run  difficulty           counts  why
ProgramBench (Almost@1) (maker)    59.0%  ██·······█·● beats 3/3    very hard · avg 10%    3.5%  at its level · small run
HealthBench Professional (maker)   62.3%  ▂·▂··▂▂▂█●▂▄ beats 10/13  hard · avg 21%         2.9%  blunt · at its level · small run
## How the score was worked out Each result implies a score on its own, after adjusting for how hard that test is and who ran it. The model's score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its *counts* column above; here it is added up by who ran the tests. [Method](../method.md) 
source                                    results  share of the score
LMArena leaderboard dataset                     2   84% ▇▇▇▇▇▇▇▇▇▇
Meta vendor-reported benchmark scores f…        4   15% ▇▇
prior: pull toward the middle                        1% ▇
weighted average                                   = 81.7
Licences: LMArena leaderboard dataset ([CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)) ## How to run it No verified API or local download yet. [interactive version](https://whatmodel.sh/model/muse-wonder-full) · [JSON](https://whatmodel.sh/api?format=json&all=1)