What are the hardware requirements for LIMO?

LIMO requires approximately 65.8 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 LIMO and how large is it?

LIMO is developed by GAIR, 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 LIMO?

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

LIMO by GAIR — VRAM 65.8GB, 32K context

Name: LIMO
Author: GAIR

LIMO is an open-source language model by GAIR. Features: 32b LLM, VRAM: 65.8GB, Context: 32K, Instruction-Based, LLM Explorer Score: 0.17.

Arxiv:2502.03387 Instruct Qwen2 Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/GAIR/LIMO

LIMO Parameters and Internals

LLM Name	LIMO
Repository 🤗	https://huggingface.co/GAIR/LIMO
Model Size	32b
Required VRAM	65.8 GB
Updated	2026-08-03
Maintainer	GAIR
Model Type	qwen2
Instruction-Based	Yes
Model Files	4.9 GB: 1-of-14 4.9 GB: 2-of-14 4.9 GB: 3-of-14 4.9 GB: 4-of-14 4.9 GB: 5-of-14 4.9 GB: 6-of-14 4.9 GB: 7-of-14 4.9 GB: 8-of-14 4.9 GB: 9-of-14 4.9 GB: 10-of-14 4.9 GB: 11-of-14 4.9 GB: 12-of-14 4.9 GB: 13-of-14 2.1 GB: 14-of-14
Model Architecture	Qwen2ForCausalLM
Context Length	32768
Model Max Length	32768
Transformers Version	4.45.2
Vocabulary Size	152064
Torch Data Type	bfloat16

Rank the LIMO 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 LIMO

Best Alternatives	Context / RAM	Downloads	Likes
...y Qwen2.5coder 32B V24.1q 200K	195K / 65.8 GB	9	2
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	28	17
...wen2.5 32B Inst BaseMerge TIES	128K / 65.8 GB	22	4
Franqwenstein 35B	128K / 69.8 GB	5	8
ELYZA Thinking 1.0 Qwen 32B	128K / 65.8 GB	23	10
Archaeo 32B KTO	128K / 65.8 GB	19	4
Hamanasu Magnum QwQ 32B	128K / 65.8 GB	295	9
Qwen2.5 32B Gokgok Step3	128K / 65.7 GB	6	0
Openthaigpt R1 32B Instruct	128K / 65.8 GB	33	8
Qwen2.5 32B Dark Days Stage2	128K / 65.8 GB	9	0

Note: green Score (e.g. "73.2") means that the model is better than GAIR/LIMO.