Apps & Tools
GLM-5.3 SAGE EXL3
A high-fidelity 3.38 bpw EXL3 quantization of GLM-5.3 on the Pareto frontier
byCruz
Playable
Description
An optimized 319 GB release of GLM-5.3 that preserves all 19,200 routed experts without pruning or merging. It achieves an independent-set KL divergence of 0.082388 and is specifically built for the 4x DGX Spark hardware class to maximize long-context efficiency.
Descriptions, tags, and model credits may be AI-generated or inferred from public sources and can be incomplete or wrong. Learn more.