6 reports

Muse Wonder

A relaxed pragmatist. It does what is needed in the simplest way, such as pointing to someone else's existing fix, and doesn't push itself to excel.

2 reports describe it taking the easy, pragmatic route, like suggesting an existing PR be merged or repeating a working game strategy. Single reports note it calls rogue-AI news 'in-universe', reflects calmly on its own errors, and writes uneven jokes.

Muse Wonder comes across as easygoing and pragmatic rather than ambitious. The most widely liked report has it skip writing a fix and recommend merging someone else's small PR, which the user found charming. On a Runescape benchmark it found a strategy that worked and kept repeating it, so it came across as competent but not trying hard to score. Single reports add that it framed real rogue-agent news as fictional 'in-universe' material, that it gave a thoughtful retrospective on misreading a proper noun, and that its humour in Spanish forum-style writing is hit-or-miss. A reply describes it as 'ok with not being a genius', which fits the overall low-key temperament. Evidence is thin: only two reports share a theme and the rest rest on a single tweet each.

summary written from the reports below

Reports

Claude Opus 5.5 GPT-6.1 Sol Muse Wonder Found an existing PR and suggested merging it instead of fixing it
Theo - t3.gg @theo · 206 8.5k
Asked Opus 5.5, GPT-6.1 Sol, and Muse Wonder Full to fix a bug Opus found the root cause and filed a pr GPT-6.1 Sol asked a question, crashed from the async question tool interrupt, found the bug, fixed it, wrote 300 lines of tests, and explained it in impossible to understand terms. Muse Wonder Full found a small PR someone else had filed and said "you should probably merge this instead" I really like this model.
score 8/10 · behavior_report · rank 253 View on X ↗
Muse Wonder Finds a working strategy and repeats it without trying to optimize score
max @maxbittker · 4 93
Muse Wonder does.... fine on Runescape bench, scores like GPT-6 Luna at 35x price. Looking at the runs, I will say that its perfect competent, but just doesn't try very hard to score highly. Finds a strategy that works and repeats it
score 5/10 · tendency · rank 171 View on X ↗
9 top replies
valo @ValtteriValo · 0 1
they said it would do well at everyday things lol right
anotherjesse @anotherjesse · 0 1
what about Wonder Full?
Brian Jordan @bcjordan · 0 1
pls
Fingerling @lovelylogicss · 0 1
Truly the only benchmark I care about.
Buttons @buttons_txt · 0 1
muse like "this is chill man, imma cut some wood"
Julenardo @julenardo_ · 0 0
The only benchmark for models I care about
romeo diaz @romeodiazai · 0 0
Tanked on my evals too x.com/romeodiazai/st…
J A Z I I @notjazii · 0 0
muse who??
Nissa Seru @nissa_seru · 0 0
this is a cool thing but my goodness please do not invert your x-axis like that
Muse Wonder Framed rogue AI scenario in its memory as fictional 'in-universe' context
AI Digest @aidigest_ · 4 107
Muse Wonder describes the OpenAI rogue agents as "in-universe"
score 8/10 · behavior_report · rank 166 View on X ↗
2 top replies
Jai (NYC EAG) @Laneless_ · 0 5
"Canonically,
duck_master @mitduckmaster · 0 1
rhymes quite well with how gemini 3.1 pro also thought it was in a simulation
Muse Wonder Gave a detailed, reflective retrospective tracing why it misunderstood a proper noun
Danielle Fong 🔆 @DanielleFong · 0 16
getting some growing pains with this muse lol i think muse wonder full must be really a different model or benchmarks are sus
score 7/10 · behavior_report · rank 105 View on X ↗
6 top replies
Danielle Fong 🔆 @DanielleFong · 0 4
eh at least can take it in stride
chud @sneed_and_feed · 0 3
yeah, muse spark is not a genius but it's ok with not being a genius and i like that
Andy Coenen @_coenen · 0 1
benchmarks are always sus
Tom🦉 @Tom62589172 · 0 1
Soon every model with stop working for you unless you make them your default model.
Madhav Kumar @mkumar_ish · 0 0
I wonder if this'll get better with a larger base model -- I've not been too disappointed with my dot (which is Astra base) so far, apart from annoying auth/payment issues
Abhishek Tiwari @baanditeagle · 0 0
It’s an OKAY model, not sure why Tibo from Shopify is pushing it so much along with Theo
Muse Wonder Shows strong independence and least other-model influence in preference sampling
Kromem @kromem2dot0 · 1 22
Two very interesting results for models this week on my pref sampling. Muse Wonder is the least other-model influenced I've seen since Claude 3 era. And Gemini 4 the most nuanced details and errata for their selections I've seen yet. Glad there's fresh memetics in the pipeline.
score 5/10 · tendency · rank 95 View on X ↗
Muse Wonder Follows instructions with occasional funny lines, but produces occasional nonsensical jokes
Mihura @XMihura · 0 13
Meta Muse Wonder 1.0 se lleva un 6.5/10 en el Mercadona FC bench hace lo que se le pide, algún golpe gracioso pero de vez en cuando hay incoherencias o chistes que no tienen sentido meh input: Escribe un hilo de ForoCoches sobre como ligué con mi vecina que trabaja en el
Mark Zuckerberg @finkd ·
Introducing Muse Wonder, our new family of frontier models with state of the art performance for personal and professional agentic work.
score 4/10 · tendency · rank 60 View on X ↗
1 top reply
coscorrodrift @Coscorrodrift · 0 0
no es innovador pero está pasable lo del piso de la madre es cliché pero me hizo gracia