Model Intelligence Sheet
ggml-org/GLM-5.3-Flash-GGUF overview
GLM 5.3 Flash Run it in the Llama app: https://llama.app Or from the terminal: bash llama serve hf ggml org/GLM 5.3 Flash GGUF Source models https://huggingfac…
Runs locally from ~9.0 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
9 GGUF files detected
Direct downloads for local inference
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| GLM-5.3-Flash-Q2_K-00001-of-00002.gguf | GGUF | Q2_K | 9.0 MB | Download |
| GLM-5.3-Flash-Q2_K-00002-of-00002.gguf | GGUF | Q2_K | 124.61 GB | Download |
| GLM-5.3-Flash-Q2_K_S-00001-of-00002.gguf | GGUF | Q2_K_S | 9.0 MB | Download |
| GLM-5.3-Flash-Q2_K_S-00002-of-00002.gguf | GGUF | Q2_K_S | 102.46 GB | Download |
| GLM-5.3-Flash-Q4_K-00001-of-00002.gguf | GGUF | Q4_K | 9.0 MB | Download |
| GLM-5.3-Flash-Q4_K-00002-of-00002.gguf | GGUF | Q4_K | 168.91 GB | Download |
| mmproj-GLM-5.3-Flash-Q8_0.gguf | GGUF | Q8_0 | 622.6 MB | Download |
| mtp-GLM-5.3-Flash-Q4_0.gguf | GGUF | Q4_0 | 4.61 GB | Download |
| mtp-GLM-5.3-Flash-Q8_0.gguf | GGUF | Q8_0 | 8.65 GB | Download |
Model Details
| Model ID | ggml-org/GLM-5.3-Flash-GGUF |
|---|---|
| Author | ggml-org |
| Pipeline | image-text-to-text |
| License | other |
| Base model | zai-org/GLM-5.3-Flash-BF16 |
| Last modified | 2026-10-19T13:40:17.000Z |
Model README
---
license: other
pipeline_tag: image-text-to-text
tags:
- gguf
- quantized
base_model:
- zai-org/GLM-5.3-Flash-BF16
---
GLM-5.3-Flash
Run it in the Llama app: https://llama.app
Or from the terminal:
llama serve -hf ggml-org/GLM-5.3-Flash-GGUFSource models
- https://huggingface.co/zai-org/GLM-5.3-Flash-BF16
Notes
- 320B MoE (18B active), 288 routed + 1 shared expert per layer, mHC + DSA, 1 MTP layer.
Q4_K: all routed experts at Q4_K.Q2_K: routed down experts at Q4_K, gate/up experts at Q2_K.Q2_K_S: all routed experts at Q2_K (smallest).- Small non-expert tensors (embeddings, attention, shared experts, dense FFN) are kept at Q8_0.
- Includes a Q8_0 mmproj for the vision encoder and MTP sidecars (Q8_0 / Q4_0) for speculative decoding.
- The Q2_K / Q2_K_S expert tensors are calibrated with the imatrix from https://huggingface.co/AesSedai/GLM-5.3-Flash-GGUF
TODOs
- add info
> [!IMPORTANT]
> This model is automatically converted using https://github.com/ggml-org/convert
Run ggml-org/GLM-5.3-Flash-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models