nvidia ai just handed dgx spark owners a real gift. nemotron labs 3 puzzle 75b a9b nvfp4 is basically built for this box. 75b total, 9.3b ac…

nvidia ai just handed dgx spark owners a real gift. nemotron labs 3 puzzle 75b a9b nvfp4 is basically built for this box. 75b total, 9.3b active, nvfp4, mamba plus moe, 256k context in the config, and it actually fits on one spark with room to breathe. on one spark, nvfp4 is the obvious move. on two sparks, you get options. keep nvfp4 and push longer context / local agents harder, or test fp8 if you want to trade memory for precision.

Source: Clem Delangue (X) | 2026-07-07