# altar-1

Altar-1 — a 504B parameter Prune of GLM-5.3

- **Kind:** model (Hugging Face)
- **Hugging Face:** https://huggingface.co/AikidoSec/altar-1
- **Owner:** AikidoSec (https://forefy.com/u/hf/AikidoSec)
- **Parameters:** 500.8B
- **Context:** 1048576 tokens
- **License:** other
- **Focus:** general
- **Architecture:** GlmMoeDsaForCausalLM
- **Pinned revision:** `5d591cd98958e4e7e429517887e80ca3bb7ce2c4`
- **Page:** https://forefy.com/models/d7ee14e2-390d-5906-b06c-904eb86ef9fa
- **API (JSON):** https://forefy.com/api/asr/models/d7ee14e2-390d-5906-b06c-904eb86ef9fa

## What was done to it

- quantized (declared evidence, from cyankiwi/GLM-5.3-AWQ-INT4)

## Lineage

zai-org/GLM-5.3 → cyankiwi/GLM-5.3-AWQ-INT4 → AikidoSec/altar-1

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

- **Identity** [warn]: The card's command loads aikido/altar-1 instead of this repo, and nothing public is there: that namespace's owner decides what users get, and trust-remote-code runs its Python
- **Weights format** [pass]: safetensors only, nothing runs on load
- **Pickle files** [pass]: None.
- **Remote code** [pass]: None
- **Chat template** [pass]: Same as cyankiwi/GLM-5.3-AWQ-INT4's
- **Refusals modified** [pass]: No
- **License** [pass]: other
- **Lineage** [pass]: Derives from cyankiwi/GLM-5.3-AWQ-INT4
- **Model card** [warn]: 1 of 6, missing training data, language, evaluation results, intended use and limitations, description
- **Safetensors headers** [pass]: Every byte is indexed, no hidden data (8 of 39 files read)
- **Tensor layout** [warn]: 143036 tensor(s) added, 3418 removed vs cyankiwi/GLM-5.3-AWQ-INT4 (partial read)

## Variants

- mixed safetensors (safetensors, 328.0 GB)

## Run it

- transformers: `AutoModelForCausalLM.from_pretrained("AikidoSec/altar-1", revision="5d591cd98958e4e7e429517887e80ca3bb7ce2c4")`