What are the hardware requirements for GRM 3.2 Sky Abliterated?

GRM 3.2 Sky Abliterated requires approximately 70.2 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed GRM 3.2 Sky Abliterated and how large is it?

GRM 3.2 Sky Abliterated is developed by Crata, a model with 35.1b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

Where can I download or evaluate GRM 3.2 Sky Abliterated?

GRM 3.2 Sky Abliterated is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

GRM 3.2 Sky Abliterated by Crata — VRAM 70.2GB

Name: GRM 3.2 Sky Abliterated
Author: Crata

GRM 3.2 Sky Abliterated is an open-source language model by Crata. Features: 35.1b LLM, VRAM: 70.2GB, License: apache-2.0, LLM Explorer Score: 0.25.

Abliterated Base model:finetune:orionllm/g... Base model:orionllm/grm-3.2-sk... Conversational Image-text-to-text Qwen3 5 moe Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/Crata/GRM-3.2-Sky-abliterated

GRM 3.2 Sky Abliterated Parameters and Internals

LLM Name	GRM 3.2 Sky Abliterated
Repository 🤗	https://huggingface.co/Crata/GRM-3.2-Sky-abliterated
Base Model(s)	OrionLLM/GRM-3.2-Sky OrionLLM/GRM-3.2-Sky
Model Size	35.1b
Required VRAM	70.2 GB
Updated	2026-10-14
Maintainer	Crata
Model Type	qwen3_5_moe
Model Files	49.7 GB: 1-of-2 20.5 GB: 2-of-2
Model Architecture	Qwen3_5MoeForConditionalGeneration
License	apache-2.0
Model Max Length	262144
Transformers Version	5.14.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Errors	replace

Rank the GRM 3.2 Sky Abliterated capabilities

Have you tried this model? Rate its performance — this feedback helps the ML community find the right model for their needs.

Instruction Following and Task Automation
Factuality and Completeness of Knowledge
Censorship and Alignment
Data Analysis and Insight Generation
Text Generation
Text Summarization and Feature Extraction
Code Generation
Multi-Language Support and Translation

Best Alternatives to GRM 3.2 Sky Abliterated

Best Alternatives	Context / RAM	Downloads	Likes
Nex N2.5 Mini	0K / 70.4 GB	3121	685
Agents A1	0K / 70.6 GB	23272	629
Nex N2 Mini	0K / 70.4 GB	2289	294
...l Nova V1.2 Experimental NVFP4	0K / 21.9 GB	161	1
Thomson 1.0 Small	0K / 70.5 GB	214	109
Huihui Agents A1 Abliterated	0K / 70.6 GB	3339	29
Occamy 1.0 FP8	0K / 36.8 GB	445	4
...yJ5ua3Sebfyr8kcVwLAdaiJLUC8pkW	0K / 70.4 GB	1894	0
Nex N2 Mini	0K / 70.4 GB	229	1
Affine 5g4yy75zuz T6	0K / 70.2 GB	756	0

Note: green Score (e.g. "73.2") means that the model is better than Crata/GRM-3.2-Sky-abliterated.