3 ms·
DGX Spark runs this sized model (I personally like qwen36moe better than gemma4moe) at speeds fast enough for interactive coding sessions. Algorithmic advances
by verdverm 5mo ago
DGX Spark runs this sized model (I personally like qwen36moe better than gemma4moe) at speeds fast enough for interactive coding sessions. Algorithmic advances like DiffusionGemma ~4x token gen speeds (https://deepmind.google/models/gemma/diffusiongemma/ https://deepmind.google/models/gemma/diffusiongemma/)