What are the hardware requirements for EEVE Korean Instruct 2.8B V1.0?

EEVE Korean Instruct 2.8B V1.0 requires approximately 5.7 GB of VRAM and supports a context window of 2K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed EEVE Korean Instruct 2.8B V1.0 and how large is it?

EEVE Korean Instruct 2.8B V1.0 is developed by yanolja, a model with 2.8b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does EEVE Korean Instruct 2.8B V1.0 perform on standard benchmarks?

EEVE Korean Instruct 2.8B V1.0 has the following published scores: MMLU 53.35. Compare against reference models on this page or on the LLM Explorer leaderboards.

EEVE Korean Instruct 2.8B V1.0 by yanolja — VRAM 5.7GB, 2K context

Name: EEVE Korean Instruct 2.8B V1.0
Author: yanolja

EEVE Korean Instruct 2.8B V1.0 is an open-source language model by yanolja. Features: 2.8b LLM, VRAM: 5.7GB, Context: 2K, License: apache-2.0, Instruction-Based, LLM Explorer Score: 0.15, Arc: 58.3, HellaSwag: 72.4, MMLU: 53.4, GSM8K: 45.1.

Arxiv:2306.02707 Arxiv:2310.01377 Arxiv:2402.14714 Autotrain compatible Base model:finetune:yanolja/ee... Base model:yanolja/eeve-korean... Conversational Endpoints compatible Generated from trainer Instruct Phi Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/yanolja/EEVE-Korean-Instruct-2.8B-v1.0

EEVE Korean Instruct 2.8B V1.0 Parameters and Internals

Model Type

text generation

Supported Languages

Korean (high proficiency)

Training Details

Data Sources:

Korean-translated version of Open-Orca/SlimOrca-Dedup, Korean-translated version of argilla/ultrafeedback-binarized-preferences-cleaned

Methodology:

Direct Preference Optimization (DPO)

Hardware Used:

Axolotl

Model Architecture:

A fine-tuned version of yanolja/EEVE-Korean-2.8B-v1.0 with extended Korean vocabulary.

Input Output

Input Format:

A chat between a curious user and an AI assistant format

Accepted Modalities:

text

Output Format:

Text - AI Assistive Response

LLM Name	EEVE Korean Instruct 2.8B V1.0
Repository 🤗	https://huggingface.co/yanolja/EEVE-Korean-Instruct-2.8B-v1.0
Base Model(s)	yanolja/EEVE-Korean-2.8B-v1.0 yanolja/EEVE-Korean-2.8B-v1.0
Model Size	2.8b
Required VRAM	5.7 GB
Updated	2025-09-02
Maintainer	yanolja
Model Type	phi
Instruction-Based	Yes
Model Files	5.0 GB: 1-of-2 0.7 GB: 2-of-2
Model Architecture	PhiForCausalLM
License	apache-2.0
Context Length	2048
Model Max Length	2048
Transformers Version	4.38.0.dev0
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	58944
Torch Data Type	bfloat16

Rank the EEVE Korean Instruct 2.8B V1.0 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 EEVE Korean Instruct 2.8B V1.0

Best Alternatives	Context / RAM	Downloads	Likes
Phi 2 Instruct Apo	2K / 5.6 GB	8	0
Phi 2 Instruct V0.1	2K / 5.6 GB	8	2
Att Model	2K / 5.7 GB	8	0
Eeve2.8 Base	2K / 5.7 GB	11	0
Eeve2.8 Ko	2K / 5.7 GB	5	0
... Instruct 2.8B V1.0 20240430 2	2K / 2.9 GB	7	0
YanoljaNEXT EEVE Instruct 2.8B	2K / 5.7 GB	63	30
Phi 2 Code Instruct	2K / 5.6 GB	7	4
Dolphin 2 6 Phi 2	0K / 5.6 GB	1160	192
Dolphin 2 6 Phi 2	0K / 5.6 GB	801	198

Note: green Score (e.g. "73.2") means that the model is better than yanolja/EEVE-Korean-Instruct-2.8B-v1.0.