# GLM-5.3

GLM-5.3 uses the same base model as GLM-5.2 — every gain comes from post-training. Compared with GLM-5.2, it is much better at complex coding and long-horizon tasks:

- **Kind:** model (Hugging Face)
- **Hugging Face:** https://huggingface.co/zai-org/GLM-5.3
- **Owner:** zai-org (https://forefy.com/u/hf/zai-org)
- **Parameters:** 753.3B
- **Context:** 1048576 tokens
- **License:** other
- **Focus:** general
- **Architecture:** GlmMoeDsaForCausalLM
- **Pinned revision:** `ba32f8c19c8738e2c57c71798c601d477a7bf844`
- **Page:** https://forefy.com/models/c823ff8c-bd01-59cf-82e2-ecb8ce225c1b
- **API (JSON):** https://forefy.com/api/asr/models/c823ff8c-bd01-59cf-82e2-ecb8ce225c1b

## Checks (repo at the pinned revision; not a behavior test)

- **Identity** [pass]: No lookalike name or conflicting provenance
- **Weights format** [pass]: safetensors only, nothing runs on load
- **Pickle files** [pass]: None.
- **Remote code** [pass]: None
- **Chat template** [info]: No parent template to compare
- **Refusals modified** [pass]: No
- **License** [pass]: other
- **Lineage** [info]: No base model declared
- **Model card** [warn]: 3 of 6, missing training data, evaluation results, intended use and limitations
- **Safetensors headers** [pass]: Every byte is indexed, no hidden data (8 of 141 files read)

## Variants

- mixed safetensors (safetensors, 755.6 GB)

## Run it

- transformers: `AutoModelForCausalLM.from_pretrained("zai-org/GLM-5.3", revision="ba32f8c19c8738e2c57c71798c601d477a7bf844")`