Z.ai · Released August 2026
GLM-5.3
What it takes to run GLM-5.3 locally: memory by quantization, the GPUs that handle it, and how fast it runs.
Updated
- Parameters
- 744B (40B active)
- Architecture
- Mixture of experts
- Context
- 1.0M tokens
- License
- GLM-5.3 custom license
- Memory at Q4_K_M
- ~466GB
What do you need to run GLM-5.3?
GLM-5.3 needs about 466GB of VRAM at Q4_K_M with 8K context (451GB for the weights plus KV cache and overhead). Because it is a mixture-of-experts model with 40B active parameters, it generates text much faster than a dense model of the same size, and it tolerates partial CPU offload better.
What to buy
GPUs for GLM-5.3
Speeds are estimates at the quantization shown with 8K context. We may earn a commission when you buy through these links. Learn more.
Memory
GLM-5.3 VRAM requirements by quantization
Lower-bit quantizations shrink the model at some cost to quality. Q4_K_M is the usual sweet spot. Totals include the KV cache for the context length shown plus runtime overhead.
| Quantization | Quality | Weights | Total @ 8K ctx | Total @ 32K ctx |
|---|---|---|---|---|
| Q2_K | Rough | 279 GB | 289 GB | 291 GB |
| Q3_K_M | Usable | 363 GB | 375 GB | 377 GB |
| Q4_K_M | RecommendedDefault | 451 GB | 466 GB | 468 GB |
| Q5_K_M | Very good | 530 GB | 547 GB | 550 GB |
| Q6_K | Near-lossless | 614 GB | 634 GB | 636 GB |
| Q8_0 | Practically lossless | 791 GB | 816 GB | 818 GB |
| BF16 | Full precision | 1,488 GB | 1,534 GB | 1,536 GB |
KV cache note: Multi-head latent attention with DeepSeek-style sparse attention. Assuming the GLM-5 layout of 78 layers, each caching a 576-value latent (512 compressed plus 64 RoPE): 78 x 576 x 2 bytes = 89,856 bytes per token, about 94GB at 1M tokens, plus a small sparse-attention indexer cache. Layer count not confirmed from the config.
Performance
How fast GLM-5.3 runs
Single-user generation speed on popular GPUs that fit the model entirely in memory, at the best quantization each can hold.
Overview
About GLM-5.3
GLM-5.3 is Z.ai's flagship, released as open weights in August 2026, and it scores 44 on the Artificial Analysis Intelligence Index v4.3, level with Kimi K3 and among the highest of any open-weight model. It is a 744B mixture-of-experts with 40B active parameters, a 1M-token context, and a thinking mode, aimed at agentic coding and long-horizon tasks. At Q4_K_M it needs roughly 465GB with 8K context, so local use realistically means a 512GB Mac Studio or a multi-GPU server; the smallest community GGUFs start above 200GB. Unlike earlier GLM releases it uses a custom license with a revenue threshold, so check the terms before commercial use.
Strengths
- +Among the strongest open-weight models for agentic coding and reasoning
- +1M-token context
- +Toggleable thinking mode
Weaknesses
- −Needs roughly 465GB at Q4_K_M; even 1-bit to 2-bit builds exceed 200GB
- −40B active parameters make it slow on unified-memory machines
- −Custom license with a revenue threshold, unlike the MIT-licensed GLM-5.3-Flash
- −Text only
Setup
How to run GLM-5.3
$ llama-server -hf unsloth/GLM-5.3-GGUF:Q4_K_M -c 8192 -ngl 99Weights: zai-org/GLM-5.3 · GGUF: unsloth/GLM-5.3-GGUF
Alternatives
Models like GLM-5.3
Frequently asked questions
How much VRAM do you need to run GLM-5.3?
About 466GB at Q4_K_M with 8K tokens of context: 451.0GB of weights, 0.7GB of KV cache, and 14.1GB of runtime overhead. Higher-precision quants and longer contexts need more; see the table above.
Is GLM-5.3 free for commercial use?
GLM-5.3 is released under the GLM-5.3 custom license. Commercial use is allowed with some conditions; read the license before shipping a product.
How long a context does GLM-5.3 support?
Up to 1,048,576 tokens (1.0M). Long contexts need a lot of extra memory for the KV cache, so most local setups run 8K to 32K.
Sources: GLM-5.3 model card (Hugging Face), Unsloth: How to run GLM-5.3 locally, runaihome: GLM-5.3 open weights, license, and GGUF sizes. Capability score basis: Artificial Analysis Intelligence Index v4.3: 44 (September 2026), level with Kimi K3 and among the highest open-weight scores; the open frontier reference on the Llama Builds scale.