What are the hardware requirements for GLM 5.3 W4AFP8?

GLM 5.3 W4AFP8 requires approximately 400 GB of VRAM and supports a context window of 1024K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed GLM 5.3 W4AFP8 and how large is it?

GLM 5.3 W4AFP8 is developed by PhalaCloud, a model with 386.1b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate GLM 5.3 W4AFP8?

GLM 5.3 W4AFP8 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

GLM 5.3 W4AFP8 by PhalaCloud — VRAM 400GB, 1024K context

Name: GLM 5.3 W4AFP8
Author: PhalaCloud

GLM 5.3 W4AFP8 is an open-source language model by PhalaCloud. Features: 386.1b LLM, VRAM: 400GB, Context: 1024K, License: mit, MoE, LLM Explorer Score: 0.34.

8-bit Base model:quantized:zai-org/g... Base model:zai-org/glm-5.3-bf1... Conversational Glm Glm moe dsa Moe Quantized Region:us Safetensors Sharded Tensorflow W4afp8

Model Card on HF 🤗: https://huggingface.co/PhalaCloud/GLM-5.3-W4AFP8

GLM 5.3 W4AFP8 Parameters and Internals

LLM Name	GLM 5.3 W4AFP8
Repository 🤗	https://huggingface.co/PhalaCloud/GLM-5.3-W4AFP8
Base Model(s)	zai-org/GLM-5.3-BF16 zai-org/GLM-5.3-BF16
Model Size	386.1b
Required VRAM	400 GB
Updated	2026-09-04
Maintainer	PhalaCloud
Model Type	glm_moe_dsa
Model Files	10.0 GB: 1-of-41 10.0 GB: 2-of-41 10.0 GB: 3-of-41 10.0 GB: 4-of-41 10.0 GB: 5-of-41 10.0 GB: 6-of-41 10.0 GB: 7-of-41 10.0 GB: 8-of-41 10.0 GB: 9-of-41 10.0 GB: 10-of-41 10.0 GB: 11-of-41 10.0 GB: 12-of-41 10.0 GB: 13-of-41 10.0 GB: 14-of-41 10.0 GB: 15-of-41 10.0 GB: 16-of-41 10.0 GB: 17-of-41 10.0 GB: 18-of-41 10.0 GB: 19-of-41 10.0 GB: 20-of-41 10.0 GB: 21-of-41 10.0 GB: 22-of-41 10.0 GB: 23-of-41 10.0 GB: 24-of-41 10.0 GB: 25-of-41 10.0 GB: 26-of-41 10.0 GB: 27-of-41 10.0 GB: 28-of-41 10.0 GB: 29-of-41 10.0 GB: 30-of-41 10.0 GB: 31-of-41 10.0 GB: 32-of-41 10.0 GB: 33-of-41 10.0 GB: 34-of-41 10.0 GB: 35-of-41 10.0 GB: 36-of-41 10.0 GB: 37-of-41 10.0 GB: 38-of-41 10.0 GB: 39-of-41 4.7 GB: 40-of-41 5.3 GB: 41-of-41
Model Architecture	GlmMoeDsaForCausalLM
License	mit
Context Length	1048576
Model Max Length	1048576
Transformers Version	5.15.0
Tokenizer Class	TokenizersBackend
Padding Token	<\|endoftext\|>
Vocabulary Size	154880

Rank the GLM 5.3 W4AFP8 capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation