---
title: GLM-5.3-SF-WFP4AFP8
canonical_url: "https://www.modelscope.cn/models/siliconflow/GLM-5.3-SF-WFP4AFP8"
md_url: "https://www.modelscope.cn/models/siliconflow/GLM-5.3-SF-WFP4AFP8.md"
repository: siliconflow/GLM-5.3-SF-WFP4AFP8
last_updated: 2026-09-01
license: other
pipeline_tag: text-generation
tasks:
  - text-generation
model_type:
  - glm_moe_dsa
architectures:
  - GlmMoeDsaForCausalLM
parameters: 406.2B
tensor_type:
  - U8
  - F32
  - BF16
  - I8
library_name:
  - safetensors
language:
  - en
  - zh
downloads: 624
stars: 0
---

# GLM-5.3-SF-WFP4AFP8

> GLM-5.3-SF-WFP4AFP8 - siliconflow 在 ModelScope 开源的模型。CrossingBits GLM-5.3 WFP4AFP8 g32 (shared experts MXFP8)

siliconflow/GLM-5.3-SF-WFP4AFP8 是 ModelScope 魔搭社区上的 406.2B 参数text-generation模型，采用 other 许可。

- **Repository**: siliconflow/GLM-5.3-SF-WFP4AFP8
- **License**: other
- **Tasks**: text-generation
- **Parameters**: 406.2B
- **Downloads**: 624
- **Stars**: 0
- **Last updated**: 2026-09-01

Source: https://www.modelscope.cn/models/siliconflow/GLM-5.3-SF-WFP4AFP8

---

# GLM-5.3

## Quant Diff

CrossingBits re-quant of official `zai-org/GLM-5.3` provider block-FP8 (`e4m3`, 128×128).

| | |
|---|---|
| Recipe | `weight_fp4_activation_fp8`, group 32, `ue8m0` / `e4m3` |
| Routed experts + dense FFN + most attn | FP4 E2M1 weights + FP8 E4M3 activations |
| `mlp.shared_experts.{gate,up,down}_proj` | MXFP8 W8A8 g32 (MegaMoE) |
| BF16 | router (`mlp.gate`), MLA absorb (`kv_b` / `k_b` / `v_b`), `indexer.weights_proj`, embed, `lm_head`, `shared_head` |
| MTP | enabled (`layers.78`) |