GLM 5.3 Flash

GLM 5.3 Flash

Zhipu · dense

Default load on 2×+ DGX Spark desks. EXL3 recipes on GitHub. No measured decode yet.

Params

unknown

Fastest decode

60.4 tok/s

2× DGX Spark · EXL3 TR3 4bpw + DFlash2 · TensorFold

Refusal

Official

Setups

4

Best reported pair · not live serving speed ↗

Run on OWterminal inference

Runs on

Quoted one-stream decode on real boxes.

View all →

Get it running

One box · EXL3 K2, one boxRecipe 2 boxes · GLM on bothRecipe 2 boxes · TensorFold, 4 streamsRecipe

More from Zhipu

GLM-5.2