Muse Spark Open: benchmark results

Provider: Meta. Released 2026-10-15. Access: Open.

Aggregate Rating 1875 ± 40, rank #22 of 1640 rated models, from 68 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Meta Muse Spark Open Card - BabyVision (Tools, Mean@5)80.2Mean accuracy (%)100
Meta Muse Spark Open Card - IntPhys2 (1012 videos)68.3Accuracy (%)100
Meta Muse Spark Open Card - OSWorld Verified (361 tasks, GUI, mean@4)80.4Mean task reward (%)100
Meta Muse Spark Open Card - SimpleVQA (Tools, Mean@5)73.6Mean accuracy (%)100
Meta Muse Spark Open Card - WorldVQA (Tools, 3000 questions)49.4Accuracy (%)100
Meta Muse Spark Open Card - ZeroBench (Tools, Pass@5)50Pass@5 (%)100
AA GDP.pdf - Legal - Criterion Pass Rate87.81Criterion Pass Rate (%)99.5
AA-Omniscience Hallucination Rate24.07Hallucination Rate (%)96.5
AA GDP.pdf - HR - Criterion Pass Rate85.57Criterion Pass Rate (%)95.9
AA-Omniscience Index - Science, Engineering & Mathematics25.8Omniscience Index95.7
AA-Omniscience Index - Software Engineering (SWE) - Rust78Omniscience Index94.5
AA-Omniscience Index - Health13.2Omniscience Index93.8

Interactive version: theaggregate.ai/model?slug=muse-spark-open · How It Works · Data refreshed daily, snapshot 2026-10-16.