Muse Spark 1.1 Benchmark Results Released

ArtificialAnlys · x · 2026-07-14

Muse Spark 1.1 scored 863 on the new agentic knowledge work benchmark AA-Briefcase, tying overall with Gemini 3.5 Flash and ranking closely with NVIDIA Nemotron 3 Ultra.

This benchmark covers a large volume of real-world task input files, requiring models to generate tables, presentations, and UI mock-ups. It uses a comprehensive three-part scoring system:

Results indicate that Muse Spark 1.1's advantages primarily stem from stronger task completion and analytical quality; its rubric pass rate is 34.5%, higher than GPT-5.5 (xhigh) and slightly lower than GLM 5.2 (max). However, it is weaker in presentation quality, with a Presentation Elo of only 432, which is 67 points lower than the previous Muse Spark version and slightly below Mistral Medium 3.5.

Related event: Meta Muse Spark 1.1 Benchmarks Released, Showing Strength in Medical and Agentic Tasks(7 posts)→

Original post →

More from Models

Models channel →