// GLM / CogVideoX (Z.ai) · Vision-language
GLM-5.3-Flash: VRAM & local deployment
320B total · 18B active (MoE) parameters. About 214 GB at 4-bit and 705 GB at BF16 with one 8K-token request.
released 2026-08-25licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | Z.ai (Zhipu) · zai-org |
|---|---|
| Family / series | GLM / CogVideoX (Z.ai) · GLM-5.3-Flash |
| Type | Vision-language (image-text-to-text) |
| Parameters | 320B total · 18B active (MoE) Source: official model card“320B total parameters and just 18B active parameters” |
| Context length | 1M tokens (1,048,576)max_position_embeddings in config.json |
| Architecture | 45 layers (11 full-attention) · MLA+HYBRID · MoE |
| Released weights | FP8 (E4M3) · 328 GBdownload |
| Licence | MIT · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-08-25 (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) | 640 GB | 705 GB | 705 GB | 706 GB | 8 × H200 (1,128GB) | — |
| FP8 / INT8 / Q8_0 | 340 GB | 375 GB | 375 GB | 376 GB | 8 × H100 (640GB) | — |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 194 GB | 214 GB | 214 GB | 215 GB | 4 × RTX PRO 6000 (384GB) | Mac 512GB |
KV cache per request: 8K 0.1 GB · 32K 0.4 GB · 128K 1.5 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.
- GLM-5.3-Flash main
- GLM-5.3-Flash-BF16 BF16
05Other GLM / CogVideoX (Z.ai) models
All 35 →06Sources
Data checked 2026-10-15. Parameters, context and licence are copied from the official repository; memory figures are our estimates.