# GLM-5.3-Flash: hardware requirements
> GLM-5.3-Flash (Z.ai) is a 320B (18B active) mixture-of-experts model with a 1.0M-token context window, released under the MIT license. It needs about 201GB of memory at Q4_K_M with 8K context.
GLM-5.3-Flash is Z.ai's August 2026 mid-size model and its first natively multimodal one: 320B total parameters with 18B active, a 1M-token context, a thinking mode, and an MIT license. It ranks just behind the largest open models on the Artificial Analysis Intelligence Index while needing well under half their memory. Most of its layers use linear attention with a fixed-size state, so long contexts are cheap. Expect about 200GB at Q4_K_M with 8K context, which means a 256GB Mac Studio or a multi-GPU server; the smallest 1-bit GGUF is 93GB and fits 128GB machines, at a clear quality cost.

## Recommended hardware
- **Multi-GPU:** 8× RTX 5090 (256GB), Q4_K_M, ~29 tokens/sec
- **Mac / unified memory:** M3 Ultra 512GB (512GB), Q6_K, ~17 tokens/sec

## Memory by quantization
| Quant | Weights | Total @ 8K ctx | Total @ 32K ctx |
| --- | --- | --- | --- |
| Q2_K | 120.0GB | 124.3GB | 124.6GB |
| Q3_K_M | 156.0GB | 161.4GB | 161.6GB |
| Q4_K_M | 194.0GB | 200.5GB | 200.8GB |
| Q5_K_M | 228.0GB | 235.5GB | 235.8GB |
| Q6_K | 264.0GB | 272.6GB | 272.9GB |
| Q8_0 | 340.0GB | 350.9GB | 351.2GB |
| BF16 | 640.0GB | 659.9GB | 660.2GB |

## GPUs that run it fully in memory (fastest first)
| GPU | Memory | Quant | Speed | Price |
| --- | --- | --- | --- | --- |
| M5 Ultra 256GB | 256GB | Q3_K_M | ~39 tok/s | ~$9,499 |
| M5 Ultra 512GB | 512GB | Q6_K | ~25 tok/s | n/a |
| M3 Ultra 512GB | 512GB | Q6_K | ~17 tok/s | $9,499 MSRP |

## How to run it
- llama.cpp: `llama-server -hf unsloth/GLM-5.3-Flash-GGUF:Q4_K_M -c 8192 -ngl 99`
- Weights: https://huggingface.co/zai-org/GLM-5.3-Flash

## Strengths
- Near-frontier coding and agentic performance with 18B active parameters
- Image input and a 1M-token context
- Small KV cache: only 11 of 45 layers keep a per-token cache
- MIT license

## Weaknesses
- About 200GB at Q4_K_M, beyond all but the largest workstations and Macs
- The official FP8 weights are far too large for home use, so local setups rely on community GGUF quants
- llama.cpp support for its hybrid attention is new
Full page: https://llamabuilds.ai/models/glm-5-3-flash · Hardware guide: https://llamabuilds.ai/models/glm-5-3-flash/requirements

---
Source: https://llamabuilds.ai/models/glm-5-3-flash · Updated October 14, 2026 · Llama Builds. Speeds marked ~ are estimates from our bandwidth model (https://llamabuilds.ai/methodology); measured results are labeled as such.