Model reference · open weights

gpt-neox-japanese

LLMs abeja Text gen 1 build Open weights 90k dl/mo

gpt-neox-japanese is an open-weight language model from abeja. gpt-neox-japanese-2.7b (BF16) weighs 5.4 GB; the smallest configuration that runs it is RTX 3060 12 GB.

For self-hosting, compare this build's requirements below with the GPU you plan to rent.

  • gpt-neox-japanese is a 2.7B-parameter text generation model developed by abeja.
  • It is designed for Japanese language tasks and supports a context length of 2048 tokens.
  • The model is released under the MIT license.

Summary of the abeja/gpt-neox-japanese-2.7b model card, 2026-10-01

What it is

Released byabeja
Released2022-08-29
VRAM5.4 GB for the weights

What it runs on

Memory and cards for gpt-neox-japanese-2.7b (BF16)

5.4 GBweights, file size
328 MBcache per 1K tokens
588 MBruntime overhead, at least
2,048 tokenscontext max
CardRequests at onceContext maxMemory
2K each
RTX 3060 12 GB8all 2K11.6 GB
RTX 4060 Ti 16 GB14all 2K15.4 GB
RTX 3090 24 GB26all 2K23.4 GB
RTX 4090 24 GB25all 2K23.4 GB
RTX 5090 32 GB37all 2K31.0 GB
L40S 48 GB56all 2K44.0 GB
A100 80 GB107all 2K78.2 GB
H100 80 GB103all 2K78.1 GB
RTX PRO 6000 Blackwell 96 GB126all 2K93.8 GB
DGX Spark (GB10) 128 GB unified146all 2K107 GB
H200 141 GB192all 2K138 GB
B200 180 GB248all 2K176 GB
Memory needed at each load
Requests at once2K tokens each
16.6 GB
59.3 GB
811.3 GB
1616.7 GB
3227.4 GB
6448.9 GB

One card, with vLLM's small-card settings.

From the model card

What abeja says about gpt-neox-japanese

Read the model card

The open PR is merged on 2022/9/14. You can use this model with v4.23 and higher versions of transformers as follows,

pip install transformers

This repository provides a 2.7B-parameter Japanese GPT-NeoX-based model. The model was trained by ABEJA, Inc

How to use

When using pipeline for text generation.

from transformers import pipeline

generator = pipeline("text-generation", model="abeja/gpt-neox-japanese-2.7b")
generated = generator(
    "人とAIが協調するためには、",
    max_length=300,
    do_sample=True,
    num_return_sequences=3,
    top_p=0.95,
    top_k=50
)
print(*generated, sep="\n")

"""
[out]
{"generated_text": "人とAIが協調するためには、「人が持っている優れた能力とAIの得意とする分野を掛け合わせる」ことが不可欠になります。"}
{"generated_text": "人とAIが協調するためには、双方の長所を活かしていくことが不可欠だと考えています。"}
{"generated_text": "人とAIが協調するためには、人間がAIを理解する、ということが重要です。人間には「AIに対してAIが何をするべきか」ということを明確に教えないと、AIはある程度の知識はあっても何をすべきかがわかりません。だから、コンピューターが考えたり、決めたりすることはAIではなく、人間が解釈して理解できるようにしなくて"}
"""

When using PyTorch.

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("abeja/gpt-neox-japanese-2.7b")
model = AutoModelForCausalLM.from_pretrained("abeja/gpt-neox-japanese-2.7b")

input_text = "人とAIが協調するためには、"
input_ids = tokenizer.encode(input_text, return_tensors="pt")
gen_tokens = model.generate(
    input_ids,
    max_length=100,
    do_sample=True,
    num_return_sequences=3,
    top_p=0.95,
    top_k=50,
)
for gen_text in tokenizer.batch_decode(gen_tokens, skip_special_tokens=True):
    print(gen_text)

Dataset

The model was trained on Japanese CC-100, Japanese Wikipedia, and Japanese OSCAR.

Tokenization

The model uses a special sub-word tokenizer. Please refer the original repository or GPT-NeoX-Japanese in detail.

Licenese

The MIT license

Quoted from the model card on Hugging Face. The full card is behind the Hugging Face link above.

© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms