What are the hardware requirements for Bilingual GPT Neox 4B Instruction Sft?

Bilingual GPT Neox 4B Instruction Sft requires approximately 7.6 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 Bilingual GPT Neox 4B Instruction Sft and how large is it?

Bilingual GPT Neox 4B Instruction Sft is developed by rinna, a model with 4b parameters. The model is published as open weights on Hugging Face and indexed on LLM Explorer with full benchmark history.

How does Bilingual GPT Neox 4B Instruction Sft perform on standard benchmarks?

Bilingual GPT Neox 4B Instruction Sft has the following published scores: MMLU 23.12. Compare against reference models on this page or on the LLM Explorer leaderboards.

Bilingual GPT Neox 4B Instruction Sft by rinna — VRAM 7.6GB, 2K context

Name: Bilingual GPT Neox 4B Instruction Sft
Author: rinna

Bilingual GPT Neox 4B Instruction Sft is an open-source language model by rinna. Features: 4b LLM, VRAM: 7.6GB, Context: 2K, License: mit, Instruction-Based, LLM Explorer Score: 0.13, Arc: 28.1, HellaSwag: 47.5, MMLU: 23.1.

Arxiv:2404.01657 Base model:finetune:rinna/bili... Base model:rinna/bilingual-gpt... Dataset:anthropic/hh-rlhf En Gpt neox Instruct Ja Pytorch Region:us Safetensors

Model Card on HF 🤗: https://huggingface.co/rinna/bilingual-gpt-neox-4b-instruction-sft

Bilingual GPT Neox 4B Instruction Sft Parameters and Internals

Model Type

bilingual, text generation, instruction-following, conversational agent

Use Cases

Areas:

Research, Commercial applications

Applications:

Language translation, Conversational agents, Text generation

Primary Use Cases:

Instruction-following conversational agent

Additional Notes

Ensure to use 'use_fast=False' for the tokenizer for correct functionality.

Supported Languages

ja (high), en (high)

Training Details

Data Sources:

Anthropic HH RLHF data, FLAN Instruction Tuning data, Japanese translations

Methodology:

Fine-tuning

Model Architecture:

36-layer, 2816-hidden-size transformer-based language model

Input Output

Input Format:

A conversation format between 'ユーザー' and 'システム', ending with 'システム: '.

Accepted Modalities:

text

Output Format:

Text response from the system in continuation of the conversation.

Performance Tips:

Adjust decoding hyper-parameters (e.g., temperature, top_p, top_k) for optimal results.

Release Notes

Version:

2023/08/02

Date:

2023-08-02

Notes:

Newly trained model with MIT license.

Version:

2023/07/31

Date:

2023-07-31

Notes:

Initial release found with non-compliant training data leading to a re-release with compliant datasets.

LLM Name	Bilingual GPT Neox 4B Instruction Sft
Repository 🤗	https://huggingface.co/rinna/bilingual-gpt-neox-4b-instruction-sft
Base Model(s)	Bilingual GPT Neox 4B rinna/bilingual-gpt-neox-4b
Model Size	4b
Required VRAM	7.6 GB
Updated	2026-05-28
Maintainer	rinna
Model Type	gpt_neox
Instruction-Based	Yes
Model Files	7.6 GB 7.6 GB
Supported Languages	ja en
Model Architecture	GPTNeoXForCausalLM
License	mit
Context Length	2048
Model Max Length	2048
Tokenizer Class	T5Tokenizer
Padding Token	[PAD]
Vocabulary Size	65536
Torch Data Type	float16

Rank the Bilingual GPT Neox 4B Instruction Sft 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 Bilingual GPT Neox 4B Instruction Sft

Best Alternatives	Context / RAM	Downloads	Likes
...al GPT Neox 4B Instruction Ppo	2K / 7.7 GB	10	14
Tora 4B	2K / 7.6 GB	7	2
...x 4B Instruction Sft En Ja 84K	2K / 7.6 GB	8	1