Artificial Analysis reported Muse Wonder Full at 57% versus Muse Spark 1.3 at 33% on Terminal-Bench 4.0. The simulated freight-dispatch example tests a tool through changing information and earlier commitments—not just a working first draft. #Muse #Meta #AI #AIBenchmarks