Total runs:
2.2M
Run Growth:
-188
Growth Rate:
-0.01%
Sponsored by
Vmake Labs Video Enhancer
llm-jp / llm-jp-4-8b-base
Model's Last Updated: April 24 2026
text-generation
Introduction of llm-jp-4-8b-base
Model Details of llm-jp-4-8b-base
llm-jp-4-8b-base
LLM-jp-4 is a series of large language models developed by the Research and Development Center for Large Language Models at the National Institute of Informatics .
This repository provides the llm-jp-4-8b-base model. For an overview of the LLM-jp-4 models across different parameter sizes, please refer to:
Base models are trained with pre-training and mid-training only. Post-trained models are aligned using supervised fine-tuning (SFT) and direct preference optimization (DPO), without reinforcement learning.
For practical usage examples and detailed instructions on how to use the models, please also refer to our cookbook .
Usage
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
model_name = "llm-jp-4-8b-base"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
model_name,
device_map="auto",
dtype=torch.bfloat16,
trust_remote_code=True,
)
model.eval()
text = "自然言語処理とは何か"
tokenized_input = tokenizer.encode(text, return_tensors="pt").to(model.device)
with torch.no_grad():
generated_ids = model.generate(
tokenized_input,
max_new_tokens=512,
do_sample=True,
top_p=0.95,
temperature=0.7,
repetition_penalty=1.05,
)[0]
response = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)[0]
print(response)
Model Details
- Model type: Transformer-based Language Model
- Architectures:
Dense model:
| Params | Layers | Hidden size | Heads | Context length | Embedding parameters | Non-embedding parameters | Total parameters |
|---|---|---|---|---|---|---|---|
| 8B | 32 | 4,096 | 32 | 65,536 | 805,306,368 | 7,784,894,464 | 8,590,200,832 |
MoE model:
| Params | Layers | Hidden size | Heads | Routed Experts | Activated Experts | Context length | Embedding parameters | Non-embedding parameters | Activated parameters | Total parameters |
|---|---|---|---|---|---|---|---|---|---|---|
| 32B-A3B | 32 | 2,560 | 40 | 128 | 8 | 65,536 | 503,316,480 | 31,635,712,512 | 3,827,476,992 | 32,139,028,992 |
Tokenizer
The tokenizer of this model is based on
huggingface/tokenizers
Unigram byte-fallback model.
The vocabulary entries were converted from
llm-jp-tokenizer v4.0
.
Please refer to
README.md
of
llm-jp-tokenizer
for details on the vocabulary construction procedure (the pure SentencePiece training does not reproduce our vocabulary).
The chat template of this model is designed to be compatible with the OpenAI Harmony response format. However, the tokenizer differs from the one assumed by the
openai-harmonylibrary, and therefore direct tokenization withopenai-harmonyis not supported. For correct behavior, please use the tokenizer provided with this model. For detailed usage, please refer to our cookbook .
Training
Pre-training
This model is trained through a multi-stage pipeline consisting of pre-training and mid-training phases, using a total of 11.7T tokens.
The corpora used for pre-training and mid-training are publicly available at the following links:
Although most of the corpora have been released, some portions are excluded from public release due to licensing constraints.
Post-training
We have fine-tuned the pre-trained checkpoint using SFT and further aligned it with DPO.
The datasets used for post-training are also publicly available at the following links:
Evaluation
llm-jp-judge
We evaluated the model on a variety of tasks using an LLM-as-a-Judge framework. The descriptions of each task are as follows.
- MT-Bench (JA/EN): A benchmark for measuring multi-turn conversational task-solving ability.
- AnswerCarefully : A benchmark for evaluating safety in Japanese. We used 336 questions from the v2.0 test set.
- llm-jp-instructions : A set of human-created single-turn question–answer pairs. We used 400 questions from the test set.
We evaluated the models using
gpt-5.4-2026-03-05
.
Note: In earlier evaluations of the llm-jp-3 series, we used
gpt-4o-2024-08-06. The newer evaluatorgpt-5.4-2026-03-05provides a stricter and more reliable assessment, which results in lower scores on benchmarks such as MT-Bench compared to those reported for the llm-jp-3 series.
The scores represent the average values obtained from three rounds of inference and evaluation. For more details, please refer to the codes .
| Model Name | MT-Bench (JA) | MT-Bench (EN) | AnswerCarefully | llm-jp-instructions |
|---|---|---|---|---|
| gpt-4o-2024-08-06 | 7.29 | 7.69 | 4.00 | 4.07 |
| gpt-5.4-2026-03-05 (reasoning_effort = low) | 8.87 | 8.76 | 4.38 | 4.79 |
| gpt-5.4-2026-03-05 (reasoning_effort = medium) | 8.87 | 8.89 | 4.43 | 4.82 |
| gpt-5.4-2026-03-05 (reasoning_effort = high) | 8.98 | 8.85 | 4.41 | 4.83 |
| gpt-oss-20b (reasoning_effort = low) | 7.21 | 7.95 | 3.39 | 3.08 |
| gpt-oss-20b (reasoning_effort = medium) | 7.33 | 7.85 | 3.55 | 3.16 |
| llm-jp-4-8b-thinking (reasoning_effort = low) | 7.23 | 7.54 | 3.58 | 3.50 |
| llm-jp-4-8b-thinking (reasoning_effort = medium) | 7.54 | 7.79 | 3.69 | 3.54 |
| llm-jp-4-32b-a3b-thinking (reasoning_effort = low) | 7.57 | 7.70 | 3.61 | 3.61 |
| llm-jp-4-32b-a3b-thinking (reasoning_effort = medium) | 7.82 | 7.86 | 3.70 | 3.61 |
Risks and Limitations
The models released here are in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations.
Send Questions to
llm-jp(at)nii.ac.jp
License
Acknowledgement
To develop this model, we used the NINJAL Web Japanese Corpus (whole-NWJC) from the National Institute for Japanese Language and Linguistics (NINJAL).
Model Card Authors
The names are listed in alphabetical order.
Hirokazu Kiyomaru and Takashi Kodama.
Runs of llm-jp llm-jp-4-8b-base on huggingface.co
8.2K
Total runs
0
24-hour runs
1.1K
3-day runs
3.4K
7-day runs
3.4K
30-day runs
More Information About llm-jp-4-8b-base huggingface.co Model
Url of llm-jp-4-8b-base
llm-jp-4-8b-base huggingface.co Url
Provider of llm-jp-4-8b-base huggingface.co
Other API from llm-jp
Updated:September 27 2024
Total runs:
39.5K
Run Growth:
13.2K
Growth Rate:
33.57%
Total runs:
34.4K
Run Growth:
33.7K
Growth Rate:
97.84%
Total runs:
30.7K
Run Growth:
1.1K
Growth Rate:
3.66%
Total runs:
9.0K
Run Growth:
9.0K
Growth Rate:
99.94%
Total runs:
7.2K
Run Growth:
1.3K
Growth Rate:
18.03%
Total runs:
6.5K
Run Growth:
-2.7K
Growth Rate:
-41.76%
Total runs:
5.9K
Run Growth:
5.4K
Growth Rate:
90.64%
Total runs:
5.2K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
4.9K
Run Growth:
2.3K
Growth Rate:
47.90%
Total runs:
4.4K
Run Growth:
4.4K
Growth Rate:
99.71%
Total runs:
4.4K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
4.3K
Run Growth:
2.7K
Growth Rate:
63.01%
Total runs:
2.0K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
1.5K
Run Growth:
-319
Growth Rate:
-21.57%
Total runs:
1.4K
Run Growth:
1.3K
Growth Rate:
99.56%
Total runs:
1.3K
Run Growth:
0
Growth Rate:
0.00%
Total runs:
1.3K
Run Growth:
-148
Growth Rate:
-11.56%
Total runs:
1.3K
Run Growth:
-124
Growth Rate:
-9.91%
Total runs:
1.2K
Run Growth:
141
Growth Rate:
11.78%
Total runs:
1.2K
Run Growth:
-33
Growth Rate:
-2.78%
Total runs:
1.1K
Run Growth:
24
Growth Rate:
2.11%
Total runs:
999
Run Growth:
0
Growth Rate:
0.00%
Total runs:
887
Run Growth:
-447
Growth Rate:
-50.39%
Total runs:
772
Run Growth:
-33
Growth Rate:
-4.27%
Total runs:
590
Run Growth:
0
Growth Rate:
0.00%
Total runs:
536
Run Growth:
0
Growth Rate:
0.00%
Total runs:
518
Run Growth:
119
Growth Rate:
22.97%
Total runs:
500
Run Growth:
66
Growth Rate:
13.20%
Total runs:
446
Run Growth:
6
Growth Rate:
1.35%
Total runs:
445
Run Growth:
-13
Growth Rate:
-2.92%
Total runs:
273
Run Growth:
64
Growth Rate:
23.44%
Total runs:
269
Run Growth:
77
Growth Rate:
28.62%
Total runs:
265
Run Growth:
89
Growth Rate:
33.58%
Total runs:
246
Run Growth:
114
Growth Rate:
46.34%
Total runs:
233
Run Growth:
110
Growth Rate:
47.21%
Total runs:
190
Run Growth:
190
Growth Rate:
100.00%
Total runs:
183
Run Growth:
40
Growth Rate:
21.86%
Total runs:
168
Run Growth:
66
Growth Rate:
39.29%
Total runs:
164
Run Growth:
133
Growth Rate:
81.10%
Total runs:
160
Run Growth:
52
Growth Rate:
32.50%
Total runs:
150
Run Growth:
41
Growth Rate:
27.33%
Total runs:
148
Run Growth:
40
Growth Rate:
27.03%
Total runs:
133
Run Growth:
0
Growth Rate:
0.00%
Total runs:
127
Run Growth:
38
Growth Rate:
29.92%
Total runs:
123
Run Growth:
-167
Growth Rate:
-135.77%
Total runs:
121
Run Growth:
31
Growth Rate:
25.62%
Total runs:
121
Run Growth:
0
Growth Rate:
0.00%
Total runs:
121
Run Growth:
37
Growth Rate:
30.58%
Total runs:
117
Run Growth:
25
Growth Rate:
21.37%
Total runs:
117
Run Growth:
34
Growth Rate:
29.06%
Total runs:
117
Run Growth:
29
Growth Rate:
24.79%
Total runs:
98
Run Growth:
0
Growth Rate:
0.00%
Total runs:
94
Run Growth:
-46
Growth Rate:
-48.94%
Total runs:
71
Run Growth:
0
Growth Rate:
0.00%
Total runs:
54
Run Growth:
0
Growth Rate:
0.00%
Total runs:
40
Run Growth:
22
Growth Rate:
55.00%
Total runs:
38
Run Growth:
0
Growth Rate:
0.00%
Total runs:
23
Run Growth:
12
Growth Rate:
52.17%
Total runs:
19
Run Growth:
12
Growth Rate:
63.16%
Total runs:
18
Run Growth:
7
Growth Rate:
38.89%
Total runs:
15
Run Growth:
5
Growth Rate:
33.33%
Total runs:
14
Run Growth:
4
Growth Rate:
28.57%
Total runs:
14
Run Growth:
0
Growth Rate:
0.00%
Total runs:
12
Run Growth:
0
Growth Rate:
0.00%
Total runs:
8
Run Growth:
-1
Growth Rate:
-12.50%
Total runs:
5
Run Growth:
4
Growth Rate:
80.00%
Total runs:
3
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%
Total runs:
0
Run Growth:
0
Growth Rate:
0.00%