tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.5

Llama 3.1 Swallow - Built with Llama

Llama 3.1 Swallow is a series of large language models (8B, 70B) that were built by continual pre-training on the Meta Llama 3.1 models. Llama 3.1 Swallow enhanced the Japanese language capabilities of the original Llama 3.1 while retaining the English language capabilities. We use approximately 200 billion tokens that were sampled from a large Japanese web corpus (Swallow Corpus Version 2), Japanese and English Wikipedia articles, and mathematical and coding contents, etc (see the Training Datasets section of the base model) for continual pre-training. The instruction-tuned models (Instruct) were built by supervised fine-tuning (SFT) on the synthetic data specially built for Japanese. See the Swallow Model Index section to find other model variants.

Note: Llama-3.1-Swallow-8B-Instruct-v0.5 model was continually pre-trained from the meta-llama/Llama-3.1-8B-Instruct and then instruction-tuned with our instruction datasets.

Release History

• June 25, 2025: Released Llama-3.1-Swallow-8B-Instruct-v0.5 and Llama-3.1-Swallow-8B-v0.5.
• March 10, 2025: Released Llama-3.3-Swallow-70B-Instruct-v0.4 and Llama-3.3-Swallow-70B-v0.4.
• December 30, 2024: Released Llama-3.1-Swallow-70B-Instruct-v0.3.
• December 23, 2024: Released Llama-3.1-Swallow-8B-Instruct-v0.3.
• November 11, 2024: Released Llama-3.1-Swallow-8B-v0.2 and Llama-3.1-Swallow-8B-Instruct-v0.2.
• October 08, 2024: Released Llama-3.1-Swallow-8B-v0.1, Llama-3.1-Swallow-8B-Instruct-v0.1, Llama-3.1-Swallow-70B-v0.1, and Llama-3.1-Swallow-70B-Instruct-v0.1.

Major Updates

This release enhances the conversation capability of Llama 3.1 Swallow. The model is trained to imitate the behavior of gemma-3-27b-it. Among all open-source LLMs with <= 8 billion parameters, Llama-3.1-Swallow-8B-Instruct-v0.5 exhibits state-of-the-art performance on Japanese MT-Bench, outperforming its predecessor, Llama-3.1-Swallow-8B-Instruct-v0.3, by 1.5 points.

Swallow Model Index

Model	Llama-3.1-Swallow-Instruct v0.5	Llama-3.1-Swallow v0.5	Llama-3.3-Swallow v0.4	Llama-3.3-Swallow-Instruct v0.4	Llama-3.1-Swallow-Instruct v0.3	Llama-3.1-Swallow-Instruct v0.2	Llama-3.1-Swallow v0.2	Llama-3.1-Swallow-Instruct v0.1	Llama-3.1-Swallow v0.1
8B	🤗 HuggingFace	🤗 HuggingFace			🤗 HuggingFace	🤗 HuggingFace	🤗 HuggingFace	🤗 HuggingFace	🤗 HuggingFace
70B			🤗 HuggingFace	🤗 HuggingFace	🤗 HuggingFace			🤗 HuggingFace	🤗 HuggingFace

The website https://swallow-llm.github.io/ provides large language models developed by the Swallow team.

Model Details

• Model type: Please refer to Llama 3.1 MODEL_CARD for details on the model architecture.
• Language(s): Japanese English
• Library: Megatron-LM, transformers
• Tokenizer: Please refer to Llama 3.1 blog for details on the tokenizer.
• Contact: swallow[at]nlp.c.titech.ac.jp

Model Performance

Japanese MT-Bench

• We report evaluation results judged by gpt-4o-2024-08-06 as below.
  • In our releases earlier than January 1, 2025, we reported scores judged by gpt-4-1106-preview. Scores reported below are thus not directly comparable with those reported in those earlier releases.

Model	coding	extraction	humanities	math	reasoning	roleplay	stem	writing	JMTAvg
llm-jp-3-7.2b-instruct3	0.358	0.597	0.812	0.386	0.438	0.766	0.622	0.721	0.588
Qwen2.5-7B-Instruct	0.599	0.741	0.719	0.637	0.541	0.744	0.624	0.713	0.665
Tanuki-8B-dpo-v1.0	0.461	0.597	0.562	0.495	0.377	0.589	0.509	0.643	0.529
Llama 3 8B Instruct	0.467	0.706	0.692	0.310	0.433	0.542	0.532	0.546	0.529
Llama 3.1 8B Instruct	0.420	0.830	0.550	0.514	0.349	0.502	0.479	0.504	0.519
Llama 3 Youko 8B Instruct	0.464	0.757	0.769	0.414	0.487	0.695	0.583	0.753	0.616
Llama-3-ELYZA-JP-8B	0.389	0.706	0.647	0.426	0.613	0.684	0.533	0.697	0.587
Llama 3 heron brain 8B v0.3	0.362	0.566	0.602	0.315	0.426	0.586	0.567	0.550	0.497
Llama 3.1 Swallow 8B Instruct v0.1	0.427	0.738	0.675	0.527	0.453	0.615	0.593	0.624	0.581
Llama 3.1 Swallow 8B Instruct v0.2	0.534	0.748	0.705	0.565	0.475	0.646	0.579	0.646	0.612
Llama 3.1 Swallow 8B Instruct v0.3	0.562	0.756	0.869	0.610	0.512	0.783	0.748	0.803	0.705
Llama 3.1 Swallow 8B Instruct v0.5	0.551	0.814	0.847	0.568	0.577	0.796	0.770	0.832	0.719

Japanese tasks

Model	JCom.	JEMHopQA	NIILC	JSQuAD	XL-Sum	MGSM	WMT20-en-ja	WMT20-ja-en	JMMLU	JHumanEval	Ja Avg
	4-shot	4-shot	4-shot	4-shot	1-shot	4-shot	4-shot	4-shot	5-shot	0-shot
	EM acc	Char-F1	Char-F1	Char-F1	ROUGE-2	EM acc	BLEU	BLEU	EM acc	pass@1
llm-jp-3-7.2b-instruct3	0.780	0.297	0.570	0.882	0.132	0.344	0.251	0.189	0.422	0.196	0.406
Qwen2.5-7B-Instruct	0.915	0.429	0.391	0.891	0.168	0.632	0.211	0.192	0.623	0.532	0.498
Tanuki-8B-dpo-v1.0	0.278	0.284	0.370	0.670	0.102	0.428	0.238	0.183	0.306	0.251	0.311
Llama 3 8B Instruct	0.880	0.417	0.385	0.891	0.126	0.424	0.214	0.202	0.468	0.296	0.430
Llama 3.1 8B Instruct	0.880	0.447	0.407	0.886	0.148	0.516	0.218	0.200	0.509	0.488	0.470
Llama 3 Youko 8B Instruct	0.921	0.481	0.517	0.899	0.209	0.472	0.256	0.191	0.469	0.262	0.468
Llama-3-ELYZA-JP-8B	0.897	0.498	0.496	0.906	0.168	0.436	0.250	0.185	0.487	0.388	0.471
Llama 3 heron brain 8B v0.3	0.923	0.493	0.569	0.906	0.218	0.456	0.277	0.217	0.499	0.318	0.488
Llama 3.1 Swallow 8B Instruct v0.1	0.924	0.587	0.574	0.917	0.138	0.508	0.282	0.228	0.530	0.366	0.505
Llama 3.1 Swallow 8B Instruct v0.2	0.929	0.560	0.599	0.915	0.137	0.528	0.288	0.227	0.550	0.408	0.514
Llama 3.1 Swallow 8B Instruct v0.3	0.924	0.528	0.583	0.896	0.191	0.532	0.281	0.229	0.544	0.394	0.510
Llama 3.1 Swallow 8B Instruct v0.5	0.937	0.511	0.606	0.900	0.174	0.604	0.293	0.230	0.581	0.496	0.533

English tasks

Model	OpenBookQA	TriviaQA	HellaSWAG	SQuAD2.0	XWINO	MMLU	GSM8K	MATH	BBH	HumanEval	En Avg
	4-shot	4-shot	4-shot	4-shot	4-shot	5-shot	4-shot	4-shot	3-shot	0-shot
	Acc	EM acc	Acc	EM acc	Acc	Acc	EM acc	CoT EM Acc	CoT EM Acc	pass@1
llm-jp-3-7.2b-instruct3	0.328	0.479	0.563	0.501	0.876	0.462	0.264	0.028	0.420	0.219	0.414
Qwen2.5-7B-Instruct	0.428	0.519	0.624	0.569	0.877	0.742	0.739	0.688	0.217	0.636	0.604
Tanuki-8B-dpo-v1.0	0.334	0.283	0.469	0.501	0.816	0.377	0.487	0.178	0.333	0.288	0.406
Llama 3 8B Instruct	0.388	0.670	0.583	0.611	0.892	0.657	0.745	0.306	0.646	0.554	0.605
Llama 3.1 8B Instruct	0.366	0.699	0.592	0.600	0.904	0.680	0.743	0.376	0.690	0.624	0.627
Llama 3 Youko 8B Instruct	0.406	0.613	0.599	0.559	0.897	0.596	0.563	0.152	0.401	0.287	0.507
Llama-3-ELYZA-JP-8B	0.318	0.551	0.523	0.600	0.882	0.587	0.558	0.164	0.321	0.449	0.495
Llama 3 heron brain 8B v0.3	0.362	0.656	0.569	0.581	0.901	0.621	0.578	0.222	0.641	0.380	0.551
Llama 3.1 Swallow 8B Instruct v0.1	0.388	0.649	0.615	0.598	0.891	0.624	0.605	0.236	0.642	0.379	0.563
Llama 3.1 Swallow 8B Instruct v0.2	0.380	0.625	0.603	0.607	0.887	0.634	0.620	0.264	0.649	0.474	0.574
Llama 3.1 Swallow 8B Instruct v0.3	0.396	0.629	0.593	0.570	0.884	0.629	0.622	0.266	0.626	0.445	0.566
Llama 3.1 Swallow 8B Instruct v0.5	0.396	0.638	0.603	0.581	0.889	0.663	0.717	0.368	0.628	0.554	0.604

Evaluation Benchmarks

Japanese MT-Bench

We used Japanese MT-Bench to assess the capabilities of multi-turn dialogue with the following settings:

• Implementation: FastChat [Zheng+, 2023] (commit #e86e70d0)
• Question: Nejumi LLM-Leaderboard NEO, mtbench_ja_question_v4
• Reference Answer: swallow-evaluation, reference answer
• Prompt for Judge: Nejumi LLM-Leaderboard NEO, mtbench_ja_prompt_v1
• Judge: gpt-4o-2024-08-06
• Scoring: Absolute scale normalized to a 0-1 range, averaged over five runs.

Japanese evaluation benchmarks

We used llm-jp-eval(v1.3.0), JP Language Model Evaluation Harness(commit #9b42d41) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows:

• Multiple-choice question answering (JCommonsenseQA [Kurihara et al., 2022])
• Open-ended question answering (JEMHopQA [Ishii et al., 2024])
• Open-ended question answering (NIILC [関根, 2003])
• Machine reading comprehension (JSQuAD [Kurihara et al., 2022])
• Automatic summarization (XL-Sum [Hasan et al., 2021])
• Machine translation (WMT2020 ja-en [Barrault et al., 2020])
• Machine translation (WMT2020 en-ja [Barrault et al., 2020])
• Arithmetic reasoning (MGSM [Shi et al., 2023])
• Academic exams (JMMLU [尹ら, 2024])
• Code generation (JHumanEval [佐藤ら, 2024])

English evaluation benchmarks

We used the Language Model Evaluation Harness(v.0.4.2) and Code Generation LM Evaluation Harness(commit #0261c52). The details are as follows:

• Multiple-choice question answering (OpenBookQA [Mihaylov et al., 2018])
• Open-ended question answering (TriviaQA [Joshi et al., 2017])
• Machine reading comprehension (SQuAD2 [Rajpurkar et al., 2018])
• Commonsense reasoning (XWINO [Tikhonov and Ryabinin, 2021])
• Natural language inference (HellaSwag [Zellers et al., 2019])
• Arithmetic reasoning (GSM8K [Cobbe et al., 2021])
• Mathematical reasoning (MATH [Hendrycks et al., 2022][Lightman et al., 2024])
• Reasoning (BBH (BIG-Bench-Hard) [Suzgun et al., 2023])
• Academic exams (MMLU [Hendrycks et al., 2021])
• Code generation (HumanEval [Chen et al., 2021])

Usage

pip install vllm

from transformers import AutoTokenizer
from vllm import LLM, SamplingParams

model_name = "tokyotech-llm/Llama-3.1-Swallow-8B-Instruct-v0.5"

tokenizer = AutoTokenizer.from_pretrained(model_name)
llm = LLM(
    model=model_name,
    tensor_parallel_size=1,
)

sampling_params = SamplingParams(
    temperature=0.6, top_p=0.9, max_tokens=512, stop="<|eot_id|>"
)


message = [
    {
        "role": "user",
        "content": "東京の紅葉した公園で、東京タワーと高層ビルを背景に、空を舞うツバメと草地に佇むラマが出会う温かな物語を書いてください。",
    },
]
prompt = tokenizer.apply_chat_template(
    message, tokenize=False, add_generation_prompt=True
)

output = llm.generate(prompt, sampling_params)

print(output[0].outputs[0].text)

Training Datasets

Instruction Tuning

The following datasets were used for the instruction tuning.

• Gemma-3-LMSYS-Chat-1M-Synth
  • Single-turn Japanese instruction dataset synthesized and derived from lmsys-chat-1m [Zhang+, ICLR24]).
  • First-turn user instructions were translated into Japanese via DeepL (machine translation), and assistant responses were generated using gemma-3-27b-it. The same model, i.e., gemma-3-27b-it served as a judge for rejection sampling (n=10). Conversations containing personally identifiable information (PII) and template-based user instructions were removed. Duplicate instructions were removed.

Risks and Limitations

The models released here are still in the early stages of our research and development and have not been tuned to ensure outputs align with human intent and safety considerations.

Acknowledgements

We thank Meta Research for releasing Llama 3.1 under a generous open license.

We received various supports, including:

• AIST project: "Research and Development of Foundation Models for Generative AI in the Physical Domain"
• NEDO project: "Development of Artificial Intelligence Application Technology to Support Judgment in Design Risk Assessment Work Based on the Perspective of Skilled Persons" (JPNP18002) of "Development of Integration Technology as the Core of Next Generation Artificial Intelligence and Robotics"
• MEXT project: "Formation of R&D center to ensure transparency and reliability of generative AI models"
• AIST program: Large Generative AI Development Support Program

License

META LLAMA 3.1 COMMUNITY LICENSE and Gemma Terms of Use

Authors

Here are the team members:

• From Okazaki Laboratory, Institute of Science Tokyo, the following members:
  • Naoaki Okazaki
  • Sakae Mizuki
  • Youmi Ma
  • Sangwhan Moon
  • Koki Maeda
  • Masanari Ohi
  • Hinari Shimada
  • Taihei Shiotani
  • Koshiro Saito
  • Tatsuya Ichinose
  • Naoya Matsushita
  • Sora Miyamoto
  • Nguyen Tien Dung
  • Yuta Katayama
• From YOKOTA Laboratory, Institute of Science Tokyo, the following members:
  • Rio Yokota
  • Kazuki Fujii
  • Taishi Nakamura
  • Takumi Okamoto
  • Ishida Shigeki
  • Masaki Kawamura
  • Yukito Tajima
• From Artificial Intelligence Research Center, AIST, Japan, the following members:
  • Hiroya Takamura

How to cite

If you find our work helpful, please feel free to cite these papers.

@inproceedings{Fujii:COLM2024,
   title={Continual Pre-Training for Cross-Lingual LLM Adaptation:
Enhancing Japanese Language Capabilities},
   author={Kazuki Fujii and Taishi Nakamura and Mengsay Loem and Hiroki
Iida and Masanari Ohi and Kakeru Hattori and Hirai Shota and Sakae
Mizuki and Rio Yokota and Naoaki Okazaki},
   booktitle="Proceedings of the First Conference on Language Modeling",
   series={COLM},
   pages="(to appear)",
   year="2024",
   month=oct,
   address={University of Pennsylvania, USA},
}

@inproceedings{Okazaki:COLM2024,
   title={Building a Large Japanese Web Corpus for Large Language Models},
   author={Naoaki Okazaki and Kakeru Hattori and Hirai Shota and Hiroki
Iida and Masanari Ohi and Kazuki Fujii and Taishi Nakamura and Mengsay
Loem and Rio Yokota and Sakae Mizuki},
   booktitle="Proceedings of the First Conference on Language Modeling",
   series={COLM},
   pages="(to appear)",
   year="2024",
   month=oct,
   address={University of Pennsylvania, USA},
}

@misc{ma:arxiv2025,
      title={Building Instruction-Tuning Datasets from Human-Written Instructions with Open-Weight Large Language Models}, 
      author={Youmi Ma and Sakae Mizuki and Kazuki Fujii and Taishi Nakamura and Masanari Ohi and Hinari Shimada and Taihei Shiotani and Koshiro Saito and Koki Maeda and Kakeru Hattori and Takumi Okamoto and Shigeki Ishida and Rio Yokota and Hiroya Takamura and Naoaki Okazaki},
      year={2025},
      eprint={2503.23714},
      archivePrefix={arXiv},
      primaryClass={cs.CL},
      url={https://arxiv.org/abs/2503.23714}, 
}

References

@misc{dubey2024llama3herdmodels,
      title={The Llama 3 Herd of Models}, 
      author={Abhimanyu Dubey and Abhinav Jauhri and Abhinav Pandey and Abhishek Kadian and Ahmad Al-Dahle and Aiesha Letman and Akhil Mathur and Alan Schelten and Amy Yang and Angela Fan et al.},
      year={2024},
      eprint={2407.21783},
      archivePrefix={arXiv},
      primaryClass={cs.AI},
      url={https://arxiv.org/abs/2407.21783}, 
}