// GLM / CogVideoX (Z.ai) · Text LLM
GLM-5: VRAM & local deployment
744B total · 40B active (MoE) parameters. About 497 GB at 4-bit and 1.64 TB at BF16 with one 8K-token request.
released 2026-02-11licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | Z.ai (Zhipu) · zai-org |
|---|---|
| Family / series | GLM / CogVideoX (Z.ai) · GLM-5 |
| Type | Text LLM (text-generation) |
| Parameters | 744B total · 40B active (MoE) Source: official model card“GLM-5 scales from 355B parameters (32B active) to 744B parameters (40B active)” |
| Context length | 198K tokens (202,752)max_position_embeddings in config.json |
| Architecture | 78 layers · MLA · MoE |
| Released weights | BF16 · 1.51 TBdownload |
| Licence | MIT · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-02-11 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 1.49 TB | 1.64 TB | 1.64 TB | 1.65 TB | 8 × B300 (2,304GB) | — |
| FP8 / INT8 / Q8_0 | 791 GB | 871 GB | 873 GB | 882 GB | 8 × H200 (1,128GB) | — |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 451 GB | 497 GB | 500 GB | 508 GB | 8 × H100 (640GB) | — |
KV cache per request: 8K 0.7 GB · 32K 2.9 GB · 128K 11.8 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- SGLang — mentioned in the official model card
- Transformers — mentioned in the official model card
- KTransformers — mentioned in the official model card
04Official repositories & quantised builds
2 official repo(s) from zai-org.
05Other GLM / CogVideoX (Z.ai) models
All 35 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| GLM-5.3 | 753.3B | 504 GB | 2026-08-25 |
| GLM-5.3-Flash | 320B / 18B | 214 GB | 2026-08-25 |
| GLM-5.2 | 753.3B | 504 GB | 2026-06-16 |
| SCAIL-2 | — | — | 2026-06-09 |
| GLM-5.1 | 753.9B | 504 GB | 2026-04-03 |
| GLM-OCR | 1.32B | 1.9 GB | 2026-01-30 |
| GLM-4.7-Flash | 31.2B / 3B | 21.8 GB | 2026-01-19 |
| GLM-Image | 6.93B | 5.1 GB | 2026-01-08 |
06Sources
Data checked 2026-08-11. Parameters, context and licence are copied from the official repository; memory figures are our estimates.