What are the hardware requirements for Llama 3 Alpha Ko 8B Instruct Marlin?

Llama 3 Alpha Ko 8B Instruct Marlin requires approximately 5.7 GB of VRAM and supports a context window of 8K tokens. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Llama 3 Alpha Ko 8B Instruct Marlin and how large is it?

Llama 3 Alpha Ko 8B Instruct Marlin is developed by allganize, a model with 8b 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 Llama 3 Alpha Ko 8B Instruct Marlin?

Llama 3 Alpha Ko 8B Instruct Marlin is hosted on Hugging Face and linked from this page. LLM Explorer also lists quantized variants and similar alternatives if available.

Llama 3 Alpha Ko 8B Instruct Marlin by allganize — VRAM 5.7GB, 8K context

Name: Llama 3 Alpha Ko 8B Instruct Marlin
Author: allganize

Llama 3 Alpha Ko 8B Instruct Marlin is an open-source language model by allganize. Features: 8b LLM, VRAM: 5.7GB, Context: 8K, License: other, Quantized, Instruction-Based, LLM Explorer Score: 0.12.

4-bit 4bit Conversational Endpoints compatible Gptq Instruct Ko Llama Quantized Region:us

Model Card on HF 🤗: https://huggingface.co/allganize/Llama-3-Alpha-Ko-8B-Instruct-marlin

Llama 3 Alpha Ko 8B Instruct Marlin Parameters and Internals

Model Type

language model

Use Cases

Areas:

language processing, logical reasoning

Additional Notes

Alpha-Instruct utilizes a community-based approach and shares insights with the community.

Supported Languages

Korean (high), English (high)

Training Details

Data Sources:

KoBEST, Haerae, Korean-Human-Judgements, Orca-Math, dpo-mix-7k

Methodology:

Evolutionary Model Merging

LLM Name	Llama 3 Alpha Ko 8B Instruct Marlin
Repository 🤗	https://huggingface.co/allganize/Llama-3-Alpha-Ko-8B-Instruct-marlin
Base Model(s)	Llama 3 Alpha Ko 8B Instruct allganize/Llama-3-Alpha-Ko-8B-Instruct
Model Size	8b
Required VRAM	5.7 GB
Updated	2026-08-06
Maintainer	allganize
Model Type	llama
Instruction-Based	Yes
Model Files	5.7 GB
Supported Languages	ko
GPTQ Quantization	Yes
Quantization Type	gptq\|4bit
Model Architecture	LlamaForCausalLM
License	other
Context Length	8192
Model Max Length	8192
Transformers Version	4.41.0
Tokenizer Class	PreTrainedTokenizerFast
Padding Token	<\|reserved_special_token_250\|>
Vocabulary Size	128256
Torch Data Type	float16

Rank the Llama 3 Alpha Ko 8B Instruct Marlin 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 Llama 3 Alpha Ko 8B Instruct Marlin

Best Alternatives	Context / RAM	Downloads	Likes
...a 3 8B Instruct 262K 4bit GPTQ	256K / 5.8 GB	5	1
... 8B Instruct 262K 4bit GPTQ 02	256K / 5.7 GB	6	0
...lama 3.1 8B Instruct GPTQ INT4	128K / 5.8 GB	9169	42
...Instruct 80K Qlora Merged GPTQ	80K / 5.8 GB	6	0
...oLeo Instruct 8B 32K V0.1 GPTQ	64K / 5.7 GB	5	0
Meta Llama 3 8B Instruct GPTQ	8K / 5.8 GB	6	0
...ta Llama 3 8B Instruct GPTQ 8B	8K / 9.2 GB	7	0
Meta Llama 3 8B Instruct GPTQ	8K / 5.7 GB	2895	7
Meta Llama 3 8B Instruct GPTQ	8K / 5.8 GB	7	1
...truct Abliterated V3 GPTQ 4bit	8K / 5.8 GB	9	0

Note: green Score (e.g. "73.2") means that the model is better than allganize/Llama-3-Alpha-Ko-8B-Instruct-marlin.