‹ PublicAI Index
The LLM benchmark aggregator.
Muse Spark
Meta
Strongest in Human preference (#15 of 354), weakest in Agents (#150 of 298). Above par in 11 of 15 scopes. Among the models it meets almost everywhere, it finishes behind Claude Fable 5 and Claude Fable 5.1 and ahead of Gemini 3.6 Flash and Grok 4.5.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Human preference65.2−3.6#15/3541/1
Human preference65.2−3.6#15/3541/1
Professional56.7−5.7#17/1941/2
Finance57.9−4.5#15/1621/1
Medical58.7−7.8#16/1491/1
Legal55.9−10#36/1731/1
Reasoning57.1−8.3#29/1951/5
Mathematics58.2−5.6#21/1521/2
Science59.5−4#29/1221/1
Knowledge53.2−9.3#58/1971/3
Academic knowledge57.6−6.3#30/1231/1
Coding46.4−19.8#110/1791/6
Agentic coding46.2−20.7#106/1731/5
Agents48.9−17.5#150/2982/6
Knowledge work48.3−22.4#109/2082/2
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for Muse Spark, left for the other.
§ 3 · Sources
Where the numbers come from
4 publications, 14 figures. Every one links to the page it was read from.
LMArena Text 1489
GDPval-AA 1006
AA-Briefcase 646
Vals · CaseLaw 63.13%Vals · LegalBench 84.22%Vals · CorpFin 65.11%Vals · TaxEval 77.68%Vals · MedCode 51.31%Vals · MedScribe 85.9%Vals · SWE-bench Verified 74.4% (Mini-SWE-agent)Vals · Vibe Code Bench 19.67% (OpenHands)Vals · GPQA Diamond 89.65%Vals · MMLU Pro 87.32%Vals · AIME 96.88%
Badge
[](https://publicai.io/model-index/m/muse-spark)