What are the hardware requirements for EEVE Korean 10.8B V1.0?

EEVE Korean 10.8B V1.0 requires approximately 21.6 GB of VRAM and supports a context window of 4K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed EEVE Korean 10.8B V1.0 and how large is it?

EEVE Korean 10.8B V1.0 is developed by yanolja, a model with 10.7b 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 EEVE Korean 10.8B V1.0?

EEVE Korean 10.8B V1.0 is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

EEVE Korean 10.8B V1.0 by yanolja — VRAM 21.6GB, 4K context

Name: EEVE Korean 10.8B V1.0
Author: yanolja

EEVE Korean 10.8B V1.0 is an open-source language model by yanolja. Features: 10.7b LLM, VRAM: 21.6GB, Context: 4K, License: apache-2.0, LLM Explorer Score: 0.14.

Arxiv:2402.14714 Autotrain compatible Base model:finetune:upstage/so... Base model:upstage/solar-10.7b... Conversational Endpoints compatible Generated from trainer Llama Region:us Safetensors Sharded Tensorflow

Model Card on HF 🤗: https://huggingface.co/yanolja/EEVE-Korean-10.8B-v1.0

EEVE Korean 10.8B V1.0 Parameters and Internals

Model Type

text generation

Use Cases

Areas:

Research, Commercial applications

Primary Use Cases:

Korean language tasks

Limitations:

The model hasn't been fine-tuned with instruction-based training.

Additional Notes

The model is fine-tuned for improved understanding of the Korean language, utilizing Korean web-crawled datasets for enhanced performance. The vocabulary was carefully expanded with new Korean tokens.

Supported Languages

Korean (Extended proficiency)

Training Details

Data Sources:

Korean web-crawled datasets available on HuggingFace

Data Volume:

100GB Korean corpus

Methodology:

Subword-based embedding with a seven-stage training process involving parameter freezing, progressively training from input embeddings to full parameters

Model Architecture:

Adapting foundational models from English to Korean with causal language modeling pre-training

LLM Name	EEVE Korean 10.8B V1.0
Repository 🤗	https://huggingface.co/yanolja/EEVE-Korean-10.8B-v1.0
Base Model(s)	upstage/SOLAR-10.7B-v1.0 upstage/SOLAR-10.7B-v1.0
Model Size	10.7b
Required VRAM	21.6 GB
Updated	2025-09-02
Maintainer	yanolja
Model Type	llama
Model Files	4.9 GB: 1-of-5 4.9 GB: 2-of-5 5.0 GB: 3-of-5 4.9 GB: 4-of-5 1.9 GB: 5-of-5
Model Architecture	LlamaForCausalLM
License	apache-2.0
Context Length	4096
Model Max Length	4096
Transformers Version	4.37.0
Tokenizer Class	LlamaTokenizer
Padding Token	</s>
Vocabulary Size	40960
Torch Data Type	bfloat16

Rank the EEVE Korean 10.8B 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

Quantized Models of the EEVE Korean 10.8B V1.0

Model	Likes	Downloads	VRAM
...orean Instruct 10.8B V1.0 Gguf	23	874	6 GB

Best Alternatives to EEVE Korean 10.8B V1.0

Best Alternatives	Context / RAM	Downloads	Likes
SOLAR 10.7B Instruct V1.0 128K	128K / 21.4 GB	63	3
SOLAR 10.7B V1.0 Instruct 16K	16K / 21.4 GB	63	2
SOLAR 10.7B V1.0 Base 16K	16K / 21.4 GB	69	0
SauerkrautLM SOLAR Instruct	8K / 21.4 GB	62	46
Fimbulvetr V2 Kuro Lotus Bf16	8K / 21.5 GB	6	4
PiVoT SOLAR 10.7B RP	8K / 21.4 GB	54	7
MetaModelv2	8K / 21.4 GB	197	0
Kazemi 1.2 Solar	8K / 21.4 GB	0	0
Bucharest 0.2	5K / 21.4 GB	6	1
Somer2	4K / 21.4 GB	10	0

Note: green Score (e.g. "73.2") means that the model is better than yanolja/EEVE-Korean-10.8B-v1.0.