GLM-5.3

Overview

Z.ai's 753B MoE flagship with ~39B active parameters and a 1M context window - the most capable open-weights coding model, with open-source SOTA on Terminal Bench 3.0 and CyberGym. The official FP8 release for vLLM and SGLang on multi-GPU server nodes.

Good for

• agentic
• chat
• coding

Recommended hardware

≈ 1024 GB VRAM to run this 753B model at the FP8 quantization.

Needs a multi-GPU server node (e.g. 8x 80 GB H100, or 4x 192 GB)

Links

• View on HuggingFace
• Run on Infersec