Muse Spark 1.2 Tops Finance Agent Benchmark at Fraction of Opus Cost

EdwardSun0909 · x · 2026-08-07

Muse Spark 1.2 is the first model to break the 60% accuracy mark on ValsAI's Finance Agent v2 benchmark, which tasks models with financial analyst duties.

The model achieves this top-tier performance at a cost of just $0.77 per test—6.7x cheaper than the previous #1 model, Opus 5 ($5.12)—while operating at twice the speed.

Related event: Muse Spark 1.2 Tops Finance Agent Benchmark(4 posts)→

Original post →

More from Models

Models channel →