Meta Muse Spark 1.3 Official BenchmarkProvider report

Name: Meta Muse Spark 1.3 Official Benchmark
Published: 2026-09-02

2026-09-02Model identity, the 1M context window, native multimodal capabilities, and API pricing come from Meta's official Muse Spark 1.3 model page. Scores come from the Meta performance image provided by the user (SHA-256: 71ba983416f90c4697f0d7071b3a747ae7e1bf3b2f86cc97acb3edc65cbfbf7b). The image contains four models and 11 benchmark rows, producing 42 available scores. Dash cells do not create scores, and max/xhigh reasoning efforts are retained in raw. The image does not state the OSWorld 2.0 task release, scoring mode, or harness, so this site keeps it as a separate metric instead of mixing it with first-attempt, partial, or strict results. The two MRCR rows are likewise captured under their original interval labels without inferring a version or needle count.

Back to data sources

Meta Muse Spark 1.3 Official BenchmarkProvider report

View source

2026-09-02Model identity, the 1M context window, native multimodal capabilities, and API pricing come from Meta's official Muse Spark 1.3 model page. Scores come from the Meta performance image provided by the user (SHA-256: 71ba983416f90c4697f0d7071b3a747ae7e1bf3b2f86cc97acb3edc65cbfbf7b). The image contains four models and 11 benchmark rows, producing 42 available scores. Dash cells do not create scores, and max/xhigh reasoning efforts are retained in raw. The image does not state the OSWorld 2.0 task release, scoring mode, or harness, so this site keeps it as a separate metric instead of mixing it with first-attempt, partial, or strict results. The two MRCR rows are likewise captured under their original interval labels without inferring a version or needle count.