№ 0088Resource
MarkTechPost: Muse Spark 1.3 benchmarks breakdown
MarkTechPost summarizes Meta's numbers: 75.4 on DeepSWE v1.1 (Opus 5 74.0, GPT-5.6 Sol 72.7), 88.8 on Terminal-Bench 2.1, and 98.1 on MRCR v2 at 512K–1M context.
The article also reports 59.4 on SWE-Atlas Codebase QnA and 66.9 on OSWorld 2.0 in max mode. It lists pricing at $1.25/$4.25 per M tokens standard and $0.10/$0.20 contributor.









ChatForm
Tgmlabs