Qwen3 Highlights

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:

Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model, ensuring optimal performance across various scenarios.
Significantly enhancement in its reasoning capabilities, surpassing previous QwQ (in thinking mode) and Qwen2.5 instruct models (in non-thinking mode) on mathematics, code generation, and commonsense logical reasoning.
Superior human preference alignment, excelling in creative writing, role-playing, multi-turn dialogues, and instruction following, to deliver a more natural, engaging, and immersive conversational experience.
Expertise in agent capabilities, enabling precise integration with external tools in both thinking and unthinking modes and achieving leading performance among open-source models in complex agent-based tasks.
Support of 100+ languages and dialects with strong capabilities for multilingual instruction following and translation.

Model Overview

Qwen3-32B has the following features:

Type: Causal Language Models
Training Stage: Pretraining & Post-training
Number of Parameters: 32.8B
Number of Paramaters (Non-Embedding): 31.2B
Number of Layers: 64
Number of Attention Heads (GQA): 64 for Q and 8 for KV
Context Length: 32,768 natively and 131,072 tokens with YaRN.

For more details, including benchmark evaluation, hardware requirements, and inference performance, please refer to our blog, GitHub, and Documentation.

Quickstart

The code of Qwen3 has been in the latest Hugging Face transformers and we advise you to use the latest version of transformers.

With transformers<4.51.0, you will encounter the following error:

KeyError: 'qwen3'

The following contains a code snippet illustrating how to use the model generate content based on given inputs.

from transformers import AutoModelForCausalLM, AutoTokenizer
model_name = "Konthee/Qwen-32B-QA-abstract"
# 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
system_prompt = """\
You are an expert at answering questions by summarizing directly from the provided source.
Use your own words to convey the relevant information from the source without copying it verbatim.
All inputs (source and question) and outputs (your answer) will be in Thai.
Ensure that your answer is concise, coherent, and accurately reflects the source content.
"""
user_prompt = """\
Source : {}

Question : {} \
"""

messages = [
    {"role": "user", "content": system_prompt}
    {"role": "user", "content": user_prompt.format(source,question)},  
]
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 (</think>)
    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)

Evaluation Results

Results retrieved from the AI Benchmark 2025 QA Leaderboard https://benchmark.ai.in.th/score/leaderboard/2025-qa

Split	BERT score	Rougel score
public	0.75582	0.33500
private	0.82055	0.44755

Data sourced directly from the leaderboard metrics

This model corresponds to team 220_อย่าคับ เจนมันเวิ่นเว้อป่าวว, which secured 1st place on both the public and private leaderboards in the 2025-QA competition

APA

AI Thailand Benchmark Programs. (2025). 2025-QA: Machine Reading Comprehension Task. Retrieved June 23, 2025, from https://benchmark.ai.in.th/task/detail/2025-qa

Authors

Konthee Boonmeeprakob ([email protected])
Pitikorn Khlaisamniang ([email protected])