What are the hardware requirements for LongWriter Zero 32B?

LongWriter Zero 32B requires approximately 65.5 GB of VRAM and supports a context window of 32K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed LongWriter Zero 32B and how large is it?

LongWriter Zero 32B is developed by THU-KEG, a model with 32b 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 LongWriter Zero 32B?

LongWriter Zero 32B is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

LongWriter Zero 32B by THU-KEG — VRAM 65.5GB, 32K context

Name: LongWriter Zero 32B
Author: THU-KEG

LongWriter Zero 32B is an open-source language model by THU-KEG. Features: 32b LLM, VRAM: 65.5GB, Context: 32K, License: apache-2.0, LLM Explorer Score: 0.2.

Arxiv:2506.18841 Base model:finetune:qwen/qwen2... Base model:qwen/qwen2.5-32b Conversational Dataset:thu-keg/longwriter-zer... En Endpoints compatible Long context Qwen2 Region:us Safetensors Sharded Tensorflow Writing Zh

Model Card on HF 🤗: https://huggingface.co/THU-KEG/LongWriter-Zero-32B

LongWriter Zero 32B Parameters and Internals

LLM Name	LongWriter Zero 32B
Repository 🤗	https://huggingface.co/THU-KEG/LongWriter-Zero-32B
Base Model(s)	Qwen/Qwen2.5-32B Qwen/Qwen2.5-32B
Model Size	32b
Required VRAM	65.5 GB
Updated	2026-08-08
Maintainer	THU-KEG
Model Type	qwen2
Model Files	3.9 GB: 1-of-17 3.9 GB: 2-of-17 3.9 GB: 3-of-17 3.9 GB: 4-of-17 3.9 GB: 5-of-17 3.9 GB: 6-of-17 3.9 GB: 7-of-17 3.9 GB: 8-of-17 3.9 GB: 9-of-17 3.9 GB: 10-of-17 3.9 GB: 11-of-17 3.9 GB: 12-of-17 3.9 GB: 13-of-17 3.9 GB: 14-of-17 3.9 GB: 15-of-17 3.9 GB: 16-of-17 3.1 GB: 17-of-17
Supported Languages	en zh
Model Architecture	Qwen2ForCausalLM
License	apache-2.0
Context Length	32768
Model Max Length	32768
Transformers Version	4.43.1
Tokenizer Class	Qwen2Tokenizer
Padding Token	<\|endoftext\|>
Vocabulary Size	152064
Torch Data Type	bfloat16
Errors	replace

Rank the LongWriter Zero 32B 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 LongWriter Zero 32B

Best Alternatives	Context / RAM	Downloads	Likes
Openbuddy Qwq 32B V24.2 200K	195K / 65.8 GB	9	3
Openbuddy Qwq 32B V24.1 200K	195K / 65.8 GB	10	3
Openbuddy Qwq 32B V25.2q 200K	195K / 65.8 GB	1	4
...y Qwen2.5coder 32B V24.1q 200K	195K / 65.8 GB	9	2
DeepSeek R1 Distill Qwen 32B	128K / 65.7 GB	841633	1584
Qwen2.5 32B	128K / 65.5 GB	49206	180
RomboUltima 32B	128K / 20.7 GB	12	6
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	16	17
Ultiima 32B	128K / 65.8 GB	11	7
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	12	4

Note: green Score (e.g. "73.2") means that the model is better than THU-KEG/LongWriter-Zero-32B.