Muse Spark 1.3 (Low): benchmark results

Provider: Meta. Released 2026-09-02. Access: API.

Aggregate Rating 1724 ± 26, rank #328 of 2286 rated models, from 21 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
ObviousBench99.31Answer pass³ (%)89.9
Thrive Helpdesk - Difficult Tickets36.54Correct ticket resolution (%)32.1
Web Clone Bench v1 - Slack4.9Full score on the Slack clone (%)23.1
Web Clone Bench v1 - Zendesk1.7Full score on the Zendesk clone (%)23.1
cua-speedrun - OSWorld (50 Tasks)67.62Mean verifier score (%), partial credit over the task set20
Bug Hunt Bench - VS Code Extension5Planted Bugs Fixed (out of 45)18.7
Web Clone Bench v1 - Sheets0.4Full score on the Sheets clone (%)14.1
Argo-Bench15.15Mean task score (0-100) over 210 enterprise data-science tas13.4
Web Clone Bench v1 - Craigslist0Full score on the Craigslist clone (%)12.8
Bug Hunt Bench9.7Planted Bugs Fixed (out of 105)11.9
Epoch AI - Cursorbench29.3Score11.8
Argo-Bench - Solved5.71Share of the 210 tasks solved (%), a task counting as solved11.6

Interactive version: theaggregate.ai/model?slug=muse-spark-1-3-low · How It Works · Data refreshed daily, snapshot 2026-11-20.