GLM-5.3-Flash requirements — can you run it?
~320.8B parameters · 12 quantizations measured · updated 2026-09-06 · source: unsloth/GLM-5.3-Flash-GGUF
Every quantization: real file size vs what you actually need
| Quant | File (measured) | VRAM/RAM needed (8K ctx) | Runs on (first fit) |
|---|---|---|---|
| UD-IQ1_S | 93.1 GB | 113.2 GB | Mac 128GB unified |
| UD-IQ1_M | 97.6 GB | 118.6 GB | Mac 256GB unified |
| UD-IQ2_XXS | 101.8 GB | 123.7 GB | Mac 256GB unified |
| UD-Q2_K_XL | 108.7 GB | 132.0 GB | Mac 256GB unified |
| UD-IQ3_XXS | 120.4 GB | 145.9 GB | Mac 256GB unified |
| UD-Q3_K_XL | 147.5 GB | 178.5 GB | Mac 256GB unified |
| UD-IQ4_XS | 156.8 GB | 189.7 GB | Mac 256GB unified |
| UD-Q4_K_XL | 199.7 GB | 241.1 GB | Mac 512GB unified |
| UD-Q5_K_XL | 240.3 GB | 289.9 GB | Mac 512GB unified |
| UD-Q6_K_XL | 291.8 GB | 351.7 GB | Mac 512GB unified |
| Q8_0 | 341.0 GB | 410.7 GB | Mac 512GB unified |
| BF16 | 641.6 GB | 771.5 GB | 1TB server RAM (CPU) |
KV cache uses a flat +20% context/overhead rule because this repo does not publish its config; exact numbers may differ for long contexts. MoE models keep all experts in memory — total size counts, not just active params.
Can I run GLM-5.3-Flash on my GPU? (UD-IQ1_S recommended quant)
| Hardware | Memory | Verdict |
|---|---|---|
| RTX 3060 12GB | 12 GB | no |
| RTX 4060 Ti 16GB | 16 GB | no |
| RTX 3090 24GB | 24 GB | no |
| RTX 4090 24GB | 24 GB | no |
| RTX 5090 32GB | 32 GB | no |
| RTX PRO 6000 96GB | 96 GB | no |
| Mac 16GB unified | 16 GB | no |
| Mac 32GB unified | 32 GB | no |
| Mac 64GB unified | 64 GB | no |
| Mac 128GB unified | 128 GB | RUNS |
| Mac 256GB unified | 256 GB | RUNS |
| Mac 512GB unified | 512 GB | RUNS |
| 32GB system RAM (CPU) | 32 GB | no |
| 64GB system RAM (CPU) | 64 GB | no |
| 128GB system RAM (CPU) | 128 GB | RUNS |
| 256GB system RAM (CPU) | 256 GB | RUNS |
| 1TB server RAM (CPU) | 1024 GB | RUNS |
| 2TB server RAM (CPU) | 2048 GB | RUNS |
How much VRAM does GLM-5.3-Flash need?
The honest answer: 113.2 GB at the UD-IQ1_S quant with 8K context — that is the measured file size (93.1 GB) plus KV cache and runtime overhead. The smallest published build needs 113.2 GB; the lossless one needs 771.5 GB.
Can I run it with CPU offload?
Yes, if combined RAM+VRAM ≥ file size — but generation speed is limited by memory bandwidth. Expect roughly: DDR5 dual channel ~10-30 tok/s for small MoE actives, single digits for big ones, 0.1-1 tok/s when experts page from disk. CPU-only is fine for batch jobs, painful for chat.
Why do other sites show different numbers?
Most pages compute weights from a formula (params × bits/8). We use the actual file sizes published in GGUF repos, which include the embedding table, unquantized tensors, and container overhead — that is why our numbers can differ from a naive calculation.
FAQ
How much VRAM does GLM-5.3-Flash need?
The measured answer: 113.2 GB at the UD-IQ1_S quant with 8K context. The smallest published build needs 113.2 GB; the lossless (F16/BF16) build needs 771.5 GB. These are real GGUF file sizes from Hugging Face, not formula estimates.
Can I run GLM-5.3-Flash on a 24GB GPU (RTX 3090/4090)?
Not at UD-IQ1_S (113.2 GB). Nothing fits 24GB — the smallest build needs 113.2 GB. Use the hosted API or a smaller model.
Can I run GLM-5.3-Flash on CPU with system RAM?
Yes, if your usable RAM is at least the file size. At UD-IQ1_S you need ~113.2 GB of RAM. CPU generation is memory-bandwidth-bound: expect single-digit tok/s for large MoE models, 10-30 tok/s for small active params, and below 1 tok/s when experts page from disk.
Why do VRAM numbers for GLM-5.3-Flash differ between sites?
Most sites compute weights from a formula (params x bits/8). ModelFit uses the actual GGUF file sizes published on Hugging Face, which include the embedding table, unquantized tensors, and container overhead — so our numbers reflect what you really download and load.