What are the hardware requirements for Ko R1 1.5B Preview?

Ko R1 1.5B Preview requires approximately 7.1 GB of VRAM and supports a context window of 128K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Ko R1 1.5B Preview and how large is it?

Ko R1 1.5B Preview is developed by OLAIR, a model with 1.5b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Ko R1 1.5B Preview perform on standard benchmarks?

Ko R1 1.5B Preview has the following published scores: LMSYS ELO 1389. Compare against reference models on this page or on the LLM Explorer leaderboards.

Ko R1 1.5B Preview by OLAIR — VRAM 7.1GB, 128K context

Name: Ko R1 1.5B Preview
Author: OLAIR

Ko R1 1.5B Preview is an open-source language model by OLAIR. Features: 1.5b LLM, VRAM: 7.1GB, Context: 128K, License: apache-2.0, LLM Explorer Score: 0.3, ELO: 1389.

Arxiv:2501.02448 Conversational Endpoints compatible Ko Qwen2 Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/OLAIR/ko-r1-1.5b-preview

Ko R1 1.5B Preview Benchmarks

LMSys ELO 1389 vs 1272 (so35) 9.2%

^nn.n% — How the model compares to the reference models: Anthropic Sonnet 3.5 ("so35"), GPT-4o ("gpt4o") or GPT-4 ("gpt4").

What is the LLM Explorer Rank (Score)

Ko R1 1.5B Preview Parameters and Internals

LLM Name	Ko R1 1.5B Preview
Repository 🤗	https://huggingface.co/OLAIR/ko-r1-1.5b-preview
Model Size	1.5b
Required VRAM	7.1 GB
Updated	2026-07-05
Maintainer	OLAIR
Model Type	qwen2
Model Files	5.0 GB: 1-of-2 2.1 GB: 2-of-2
Supported Languages	ko
Model Architecture	Qwen2ForCausalLM
License	apache-2.0
Context Length	131072
Model Max Length	131072
Transformers Version	4.48.1
Tokenizer Class	LlamaTokenizer
Padding Token	<｜end▁of▁sentence｜>
Vocabulary Size	151665
Torch Data Type	float32

Rank the Ko R1 1.5B Preview 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 Ko R1 1.5B Preview

Best Alternatives	Context / RAM	Downloads	Likes
Qwen2.5 1.5B 1.3M Stretched	1280K / 3.1 GB	11	0
ReaderLM V2	500K / 3.1 GB	35356	804
Reader Lm 1.5B	250K / 3.1 GB	385	608
DeepSeek R1 Distill Qwen 1.5B	128K / 3.5 GB	600051	1545
Qwen2.5 1.5B	128K / 3.1 GB	863295	203
VibeThinker 1.5B	128K / 3.5 GB	3417	552
Merge Cord V1	128K / 3.1 GB	155	1
...FT Tulu3 Decontaminated Masked	128K / 3.1 GB	167	0
Qwen2 1.5B	128K / 3.1 GB	195380	102
Seq Monkey Pretrain Base 1.5B	128K / 3.5 GB	37	0

Note: green Score (e.g. "73.2") means that the model is better than OLAIR/ko-r1-1.5b-preview.