What are the hardware requirements for Kogpt?

Kogpt requires approximately 1.6 GB of VRAM. Quantized variants may run on less VRAM; see the Quantized Models section on this page.

Who developed Kogpt and how large is it?

Kogpt is developed by psyche, a model with 392.3m parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Kogpt perform on standard benchmarks?

Kogpt has the following published scores: MMLU 26.56. Compare against reference models on this page or on the LLM Explorer leaderboards.

Kogpt by psyche — VRAM 1.6GB

Name: Kogpt
Author: psyche

Kogpt is an open-source language model by psyche. Features: 392.3m LLM, VRAM: 1.6GB, License: apache-2.0, LLM Explorer Score: 0.12, Arc: 21.2, HellaSwag: 28.1, MMLU: 26.6.

En Endpoints compatible Generation Gpt2 Ko Pytorch Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/psyche/kogpt

Kogpt Parameters and Internals

Model Type

text generation

Supported Languages

Korean (high)

Training Details

Context Length:

2048

Model Architecture:

Expanded GPT-2 architecture with context and embedding dimension increase.

LLM Name	Kogpt
Repository 🤗	https://huggingface.co/psyche/kogpt
Model Size	392.3m
Required VRAM	1.6 GB
Updated	2026-08-08
Maintainer	psyche
Model Type	gpt2
Model Files	1.6 GB 1.6 GB
Supported Languages	en ko
Model Architecture	GPT2LMHeadModel
License	apache-2.0
Transformers Version	4.31.0
Tokenizer Class	GPT2Tokenizer
Vocabulary Size	32002
Torch Data Type	float32
Activation Function	gelu_new

Rank the Kogpt 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