GLM-5.3

Open Source

Open Weights

Active Parameters

753.9B

Context Length

1.05M

Modality

Text

Architecture

Mixture of Experts (MoE)

License

other

Release Date

25 Aug 2026

Knowledge Cutoff

API Pricing (per 1M)

Input: $1.40 · Output: $4.40

System Requirements

VRAM requirements for different quantization methods and context sizes

1,024 tokens

1588.81 GB VRAM

Consumer

99x RTX 4090

24GB VRAM

Datacenter

25x NVIDIA A100

80GB VRAM

Apple Silicon

21x Apple M3 Max

128GB VRAM

1,048,576 tokens

5805.78 GB VRAM

Consumer

517x RTX 4090

24GB VRAM

Datacenter

111x NVIDIA A100

80GB VRAM

Apple Silicon

120x Apple M3 Max

128GB VRAM

Architecture Diagram

Evaluation Benchmarks

Rank

#10

Benchmark	Score	Rank
Agentic Index Artificial Analysis	max 0.53	6
Agentic Coding LiveBench Agentic	0.61	11
Web Development WebDev Arena	max 1614	16
Intelligence Index Artificial Analysis	max 0.45	16
General Text Text Arena	max 1483	17
Coding LiveBench Coding	0.79	18
Agent Arena Agent Arena	max 0.03	22
LiveBench Average LiveBench Average	0.76	22
General LiveBench Global	0.76	23
Coding Index Artificial Analysis	max 0.75	24
Reasoning LiveBench Reasoning	0.86	26
Mathematics LiveBench Mathematics	0.88	32
Data Analysis LiveBench Data Analysis	0.70	37

Rankings

Overall Rank

#10

Coding Rank

#11

About GLM-5.3

GLM-5.3 is a large-scale mixture-of-experts language model developed by Zhipu AI / Z.ai utilizing dynamic sparse attention. It provides state-of-the-art multilingual comprehension, complex mathematical reasoning, and coding capabilities.

Technical Specifications

Attention

Attention Structure

DeepSeek Sparse Attention

Attention Heads

Key-Value Heads

Attention Head Dimension

192

Position Embedding

ROPE

RoPE Theta

8,000,000

Sliding Window Attention

Sliding Window Size

Sliding Window Ratio

Linear Attention

Linear Attention Ratio

Normalization

RMS Normalization

Activation Function

SwigLU

Dimensions

Auxiliary Parameters

Hidden Dimension Size

6,144

Number of Layers

FFN Intermediate Size (Dense)

12,288

Multi-Token Prediction Heads

Tokenizer

Vocabulary Size

154,880

Mixture of Experts

Total Expert Parameters

51.6B

Number of Experts

256

Active Experts

Shared Experts

FFN Intermediate Size (per Expert)

2,048

Dense Layers Before MoE

Resources

Download Weights

About GLM Family

General Language Models from Z.ai