EXAONE-Deep 7.8B on WebGPU

LG AI Research's deep reasoning model. ~4.7 GB Q4_K_M. Runs entirely in your browser via WebGPU.

Click Load to start

EXAONE-Deep uses <thought>...</thought> for chain-of-thought reasoning. First WebGPU package of this model. Built for AMD Strix Halo unified memory.