Meta has released Muse Spark 1.2, its third model launch in four months, scoring 54 on the Artificial Analysis Intelligence Index, a benchmark aggregating results across nine evaluations. The score ties for third place among US labs with SpaceXAI’s Grok 4.5, behind Claude Opus 5, Claude Fable 5, GPT-5.6 Sol, and Kimi K3. The release follows a turbulent period for Meta’s AI division, which had been sidetracked after its Llama 4 launch was criticized for using different model versions for different tests, with former Chief Scientist Yann LeCun saying the team had fudged numbers. Meta then acquired a large stake in Scale AI, brought in CEO Alexandr Wang to run Meta Superintelligence Labs, and laid off hundreds of engineers. Muse Spark 1.2 arrives about a month after Muse Spark 1.1, which scored 51, and follows Muse Spark 1.0’s debut score of 43 in April, marking a jump from 43 to 54 in under four months. The 3-point gain over 1.1 is concentrated in agentic capability; on GDPval-AA v2, its Elo rating jumped 260 points from 1371 to 1631, placing fifth among tested models. Terminal-Bench 2.1 rose two points to 80 percent, and Tau3-Bench Banking gained two points to