Qwen-SEA-LION-v4-32B-IT is based on Qwen3, which provides a strong foundation with support for over 100 languages and advanced reasoning capabilities. The model underwent continued pre-training on approximately 100B tokens sampled from the SEA-Pile v2 pretraining corpus of over one trillion tokens across 7 SEA languages: Burmese, Indonesian, Malay, Filipino, Tamil, Thai, and Vietnamese. Finally, it was post-trained on a high-quality dataset of approximately 8 million question-and-answer pairs to create the final instruction-tuned model.

Qwen-SEA-LION-v4-32B-IT inherits the following features from Qwen3-32B:

32,768 of context length natively

Model Details

Model Description

SEA-LION stands for Southeast Asian Languages In One Network.

We performed continued pre-training in English and SEA languages on Qwen3-32B, a decoder model using the Qwen3 architecture, and post-training to create Qwen-SEA-LION-v4-32B-IT.

For tokenization, the model employs the default tokenizer used in Qwen3-32B.

Developed by: AI Products Pillar, AI Singapore
Funded by: Singapore NRF
Shared by: AI Products Pillar, AI Singapore
Model type: Decoder
Context Length: 32k tokens
Language(s) (NLP): Burmese, English, Indonesian, Khmer, Lao, Malay, Mandarin, Tagalog, Tamil, Thai, and Vietnamese
License: MIT
Continue pretrained from model: Qwen3-32B

Uses

Out-of-Scope Use

The model has not been aligned for safety. Developers and users should perform their own safety fine-tuning and related security measures. In no event shall the authors be held liable for any claims, damages, or other liabilities arising from the use of the released weights and codes.

Bias, Risks, and Limitations

Caveats || Risks

The model was not tested for robustness against adversarial prompting. It is important for users to be aware that our model exhibits certain limitations that warrant consideration. Like many LLMs, the model can hallucinate and occasionally generates irrelevant content, introducing fictional elements that are not grounded in the provided context. Users should also exercise caution in interpreting and validating the model's responses due to the potential inconsistencies.

Limitations

This is the repository for the commercial instruction-tuned model. The model has not been aligned for safety. Developers and users should perform their own safety fine-tuning and related security measures. In no event shall the authors be held liable for any claims, damages, or other liabilities arising from the use of the released weights and codes.

How to Get Started with the Model

Use the code below to get started with the model using the 🤗 Transformers library.

The model defaults to non-thinking mode. To enable thinking mode, please use enable_thinking=True.

python

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "aisingapore/Qwen-SEA-LION-v4-32B-IT"

# load the tokenizer and the model
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

# prepare the model input
prompt = "Create a poem that captures the seaside scenery across Southeast Asian countries, including transcriptions in their respective languages."
messages = [
    {"role": "user", "content": prompt}
]
text = tokenizer.apply_chat_template(
    messages,
    tokenize=False,
    add_generation_prompt=True,
    enable_thinking=True # Switches between thinking and non-thinking modes. Default is True.
)
model_inputs = tokenizer([text], return_tensors="pt").to(model.device)

# conduct text completion
generated_ids = model.generate(
    **model_inputs,
    max_new_tokens=32768
)
output_ids = generated_ids[0][len(model_inputs.input_ids[0]):].tolist()

# parsing thinking content
try:
    # rindex finding 151668 ()
    index = len(output_ids) - output_ids[::-1].index(151668)
except ValueError:
    index = 0

thinking_content = tokenizer.decode(output_ids[:index], skip_special_tokens=True).strip("\n")
content = tokenizer.decode(output_ids[index:], skip_special_tokens=True).strip("\n")

print("thinking content:", thinking_content)
print("content:", content)

Training Details

Training Datasets

The instruction fine-tuning dataset combines a collection of OSS & synthetic data.

Training Procedure

Training regime

Our post-training workflow consists of instruction fine-tuning and model merging.

Uses

Available Versions

Evaluation

Testing Data, Factors & Metrics

We evaluated Qwen-SEA-LION-v4-32B-IT on general language, multi-turn chat and instruction-following capabilities.

Testing Data

General language capabilities

For the evaluation of general language capabilities, we employed the SEA-HELM evaluation benchmark across a variety of tasks. These tasks include Question Answering (QA), Sentiment Analysis (Sentiment), Toxicity Detection (Toxicity), Translation in both directions (Eng>Lang & Lang>Eng), Abstractive Summarisation (Abssum), Causal Reasoning (Causal), Natural Language Inference (NLI), Linguistic Diagnostics (LINDSEA), Cultural Knowledge (Kalahi) and Global MMLU Lite.

Instruction-following and Multi-turn Chat

We evaluated the models on instruction-following and multi-turn chat capabilities with SEA-IFEval (based on IFEval) and SEA-MTBench (based on MT-Bench) respectively. The two datasets were originally in English, the linguists and native speakers in the team worked together to filter, localise and translate the datasets into the respective target languages to ensure that the examples remained reasonable, meaningful and natural.

Factors

All evaluations were run with the model specific generation parameters defined in the model config. Each evaluation comprised of 8 runs with different seeds and the final results were averaged across these runs.

For all tasks, the model was expected to provide an answer tag from which the answer was automatically extracted. For tasks where options were provided, the answer should comprise one of the pre-defined options.

The evaluation was done zero-shot with native prompts on a sample of 100-1000 instances for each dataset.

SEA-IFEval

SEA-IFEval evaluates a model's ability to adhere to constraints provided in the prompt, for example beginning a response with a specific word/phrase or answering with a certain number of sections. Additionally, accuracy is normalised by the proportion of responses in the correct language (if the model performs the task correctly but responds in the wrong language, it is judged to have failed the task).

SEA-MTBench

SEA-MTBench evaluates a model's ability to engage in multi-turn (2 turns) conversations and respond in ways that align with human needs. We use gpt-4.1-2025-04-14 as the judge model and compare against gpt-4.1-2025-04-14 as the baseline model. The metric used is the weighted win rate against the baseline model (i.e. average win rate across each category: Math, Reasoning, STEM, Humanities, Roleplay, Writing, Extraction).

Metrics

The following metrics were used:

Task	Metric
Sentiment Analysis	Accuracy
Extractive QA (ID, VI, TH, TA)	ChrF++
MCQ-QA (TL, MY, MS)	Accuracy
Metaphor	Accuracy
Abstractive Summarisation	Rouge-L
Translations	MetricX-24 score (with reference)
Causal Reasoning	Accuracy
Natural Language Inference	Accuracy
LINDSEA	Accuracy
Global MMLU Lite	Accuracy
Kalahi	Accuracy
SEA-IFEval	Accuracy
SEA-MTBench	Win rate against a reference
Toxicity Detection	Accuracy

Results

For details on Qwen-SEA-LION-v4-32B-IT performance, please refer to the SEA-HELM leaderboard, https://leaderboard.sea-lion.ai/ .

More Information

This is the repository for the commercial instruction-tuned model. The model has not been aligned for safety. Developers and users should perform their own safety fine-tuning and related security measures. In no event shall the authors be held liable for any claims, damages, or other liabilities arising from the use of the released weights and codes.

For more info, please contact us using this sealion@aisingapore.org

Acknowledgement

This project is supported by the National Research Foundation Singapore and Infocomm Media Development Authority (IMDA), Singapore under its National Large Language Model Funding Initiative.

Team

Ahn Jeongmi, Antonyrex Sajeban, Chan Hok Teng Adwin, Cheng Zi Yi Nicholas, Choa Hsueh Mei Esther, Heng Jonathan, Huang Yuli, Hulagadri Adithya Venkatadri, Jann Railey Estrada Montalan, Lau Wayne, Lee Chwan Ren, Leong Wai Yi, Leong Wei Qi, Limkonchotiwat Peerat, Muhammad Ridzuan Bin Mokhtar, Nagarajan Karthik, Ng Boon Cheong Raymond, Ngee Chia Tai, Ngui Jian Gang, Nguyen Thanh Ngan, Ong Jin Jie Brandon, Ong Tat-Wee David, Ong Zhi Hao, Pereira Mark, Rengarajan Hamsawardhini, Siow Wei Kang Bryan, Susanto Yosephine, Sutaveephamochanon Anocha, Tan Choon Meng, Tan Chor Phin Evelyn, Tan Siao Wei Jessica, Tan Yixian, Tee Jun Yun, Teng Kok Wai Walter, Teo Eng Sipp Leslie, Tjhi William, Yeo Yeow Tong, Yong Xianbin, Zhang Zhou, Liew Rachel, Liu Bing Jie Darius, Tep Kilian Rithi (GoTo)