What are the hardware requirements for PULI GPTrio?

PULI GPTrio requires approximately 15.4 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 PULI GPTrio and how large is it?

PULI GPTrio is developed by NYTK, a model with 7.7b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does PULI GPTrio perform on standard benchmarks?

PULI GPTrio has the following published scores: MMLU 24.73. Compare against reference models on this page or on the LLM Explorer leaderboards.

PULI GPTrio by NYTK — VRAM 15.4GB, 2K context

Name: PULI GPTrio
Author: NYTK

PULI GPTrio is an open-source language model by NYTK. Features: 7.7b LLM, VRAM: 15.4GB, Context: 2K, License: cc-by-nc-4.0, LLM Explorer Score: 0.13, Arc: 30.7, HellaSwag: 53.5, MMLU: 24.7, GSM8K: 0.8.

En Endpoints compatible Gpt neox Hu Puli Pytorch Region:us Safetensors Sharded Tensorflow Zh

Model Card on HF 🤗: https://huggingface.co/NYTK/PULI-GPTrio

PULI GPTrio Parameters and Internals

Model Type

text-generation, trilingual

Use Cases

Areas:

research, multilingual applications

Limitations:

max_seq_length = 2048, float16, vocab size: 150 016

Supported Languages

hu (native), en (native), zh (native)

Training Details

Data Sources:

Hungarian corpus, English corpus, Github, Chinese corpus

Data Volume:

Hungarian: 41.5B words (314 GB), English: 61.9B words (391 GB), Chinese: 98.7B Chinese characters (340 GB), 6 million Github documents (33 GB)

Methodology:

Trained with EleutherAI's GPT-NeoX

Context Length:

2048

Model Architecture:

GPT-NeoX

LLM Name	PULI GPTrio
Repository 🤗	https://huggingface.co/NYTK/PULI-GPTrio
Model Size	7.7b
Required VRAM	15.4 GB
Updated	2026-07-18
Maintainer	NYTK
Model Type	gpt_neox
Model Files	5.0 GB: 1-of-4 5.0 GB: 2-of-4 4.2 GB: 3-of-4 1.2 GB: 4-of-4 9.9 GB: 1-of-2 5.6 GB: 2-of-2
Supported Languages	hu en zh
Model Architecture	GPTNeoXForCausalLM
License	cc-by-nc-4.0
Context Length	2048
Model Max Length	2048
Transformers Version	4.37.0
Tokenizer Class	GPTNeoXTokenizer
Beginning of Sentence Token	<\|endoftext\|>
End of Sentence Token	<\|endoftext\|>
Unk Token	<\|endoftext\|>
Vocabulary Size	150016
Torch Data Type	float16
Errors	replace

Rank the PULI GPTrio 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