GLM 5.2 W4AFP8 — Hardware Requirements & GPU Compatibility
ChatReasoning
GLM 5.2 W4AFP8 is a 391.9B-parameter open language model from PhalaCloud in the GLM 5 family. It supports a context window of up to 1,048,576 tokens. At Q4_K_M it needs about 239.34 GB of VRAM — see which GPUs and Macs can run it below.
GLM 5.2 W4AFP8 runs on a 256 GB+ Mac or a datacenter GPU
Best local pick → Mac Studio (M3 Ultra, 256GB) · Q4_K_M · 236.9 GB · ~2 tok/s
Macs from 192 GB (Q2_K · ~3 tok/s)
MIT
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run GLM 5.2 W4AFP8
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve PhalaCloud/GLM-5.2-W4AFP8 --max-model-len 16384 --tensor-parallel-size 8vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on B300 (288 GB) (51 GB headroom) · Q4_K_M
- Generation speed
- ~22 tok/s
- generation speed
- Cost per 1M output tokens
- $88.84
- per 1M output tokens
How Much VRAM Does GLM 5.2 W4AFP8 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 170.8 GB | 169.8 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on B200 192 GB | |
| 195.3 GB | 194.3 GB | 3-bit medium — 3-bit medium quantization | Runs on Instinct MI325X 256 GB | |
| 239.3 GB | 238.4 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on Instinct MI325X 256 GB | |
| 283.4 GB | 282.5 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on Instinct MI350X 288 GB | |
| 327.5 GB | 326.5 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on 5× H100 SXM 80 GB | |
| 396.1 GB | 395.1 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 6× H100 SXM 80 GB | |
| 787.9 GB | 787 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 7× H200 SXM 141 GB |
Selected Q4_K_M
4-bit medium quantization236.9GBat 16K context
- Weights
- 235.1 GB
- KV cache
- 1.5 GB
- Overhead
- 0.3 GB
Server-class: ~236.9 GB at 16K context · one Instinct MI325X 256 GB
Which Hardware Can Run GLM 5.2 W4AFP8?
Q4_K_M · 239.3 GBGLM 5.2 W4AFP8 (Q4_K_M) requires 239.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 312+ GB is recommended. Using the full 1049K context window can add up to 90.3 GB, bringing total usage to 329.6 GB. No consumer GPU has enough memory. 2 devices with unified memory can run GLM 5.2 W4AFP8, including Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 256GB).
192+ GB
Q4_K_M · best fit Q2_K (168.3 GB)
6 GPUs · 1 Mac class · 2 device families
B200~31Simulate output on B200B300~31Simulate output on B300Instinct MI350X~28Simulate output on Instinct MI350XM3 Ultra256–512 GB~3Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~102Simulate output on NVIDIA DGX H1008x A100 SXM4~62Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 128 GB: too small for this model (needs ≥ 168.3 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
Alternatives to GLM 5.2 W4AFP8
GLM 5 family · 6 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.5 397B A17BAlibaba | 113.0 | 242.6 GB | —needs > 24 GB | same size |
| Llama 4 Maverick 17B 128E InstructMeta | 87.1 | 272.7 GB | —needs > 24 GB | same size |
| Llama 3.1 405B InstructMeta | 83.2 | 275.6 GB | —needs > 24 GB | same size |
| Intern S2 397BInternLM | — | 242.6 GB | —needs > 24 GB | same size |
| Ornith 1.5 397Bornith-ai | — | 242.6 GB | —needs > 24 GB | same size |
| Trinity Large ThinkingArcee AI | — | 799.6 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- PhalaCloud
- Family
- GLM 5
- Parameters
- 391.9B
- Architecture
- GlmMoeDsaForCausalLM
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 154,880
- Release Date
- 2026-06-18
- License
- MIT
Frequently Asked Questions
How much VRAM does GLM 5.2 W4AFP8 need?
GLM 5.2 W4AFP8 requires 239.3 GB of VRAM at Q4_K_M, or 787.9 GB at BF16. Full 1049K context adds up to 90.3 GB (329.6 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 391.9B × 4.8 bits ÷ 8 = 235.1 GB
KV Cache + Overhead ≈ 4.2 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 94.5 GB (at full 1049K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run GLM 5.2 W4AFP8?
No — GLM 5.2 W4AFP8 requires at least 170.8 GB at Q2_K, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for GLM 5.2 W4AFP8?
For GLM 5.2 W4AFP8, Q4_K_M (239.3 GB) offers the best balance of quality and VRAM usage. Q5_K_M (283.4 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 170.8 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run GLM 5.2 W4AFP8 on a Mac?
Yes — Mac Pro M2 Ultra (192 GB) and 3 other Macs can run GLM 5.2 W4AFP8, the smallest at Q2_K. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 239.3 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 256GB).
Can I run GLM 5.2 W4AFP8 locally?
Yes — GLM 5.2 W4AFP8 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is GLM 5.2 W4AFP8?
At Q4_K_M, GLM 5.2 W4AFP8 can reach ~20 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: NVIDIA B300 → 8000 ÷ 239.3 × 0.65 = ~22 tok/s
Estimated speed at Q4_K_M (239.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of GLM 5.2 W4AFP8?
At Q4_K_M, the download is about 235.12 GB. The full-precision BF16 version is 783.72 GB. The smallest option (Q2_K) is 166.54 GB.
Which GPUs can run GLM 5.2 W4AFP8?
No single consumer GPU has enough VRAM to run GLM 5.2 W4AFP8 at Q4_K_M (239.3 GB). Multi-GPU or professional hardware is required.
Which devices can run GLM 5.2 W4AFP8?
2 devices with unified memory can run GLM 5.2 W4AFP8 at Q4_K_M (239.3 GB), including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.