Meta's New Model Tops Three Rankings in a Row, Legal Agent Overtakes Grok 4.5

According to Insight Beat, AI assessment firm Vals AI has released the Muse Spark 1.1 scores. The model ranked first in three assessments: Medical Records, Tax Q&A, and Legal Agent, with scores of 88.89%, 79.72%, and 20.00%, respectively.

In the Tax assessment TaxEval v2, it outperformed Fable 5, Opus 4.8, and GPT-5.5. The assessment included over 1200 undisclosed test questions that evaluated both answers and reasoning processes. The top models had a small performance gap, with the top ten differing by less than 3 percentage points.

In the Legal Agent assessment, Muse Spark 1.1 scored 20.00%, surpassing Grok 4.5 at 12.92% and Fable 5 at 11.25%. The test required the model to perform real legal work using documents, tables, and file tools.

It scored 72.16% in the Web Application Programming assessment, surpassing GPT-5.5, Grok 4.5, and GPT-5.4, but still lagging behind Fable 5 and several Claude models.