Instructions to use abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1") model = AutoModelForCausalLM.from_pretrained("abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1
- SGLang
How to use abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1 with Docker Model Runner:
docker model run hf.co/abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1
# Load model directly
from transformers import AutoTokenizer, AutoModelForCausalLM
tokenizer = AutoTokenizer.from_pretrained("abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1")
model = AutoModelForCausalLM.from_pretrained("abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1", device_map="auto")
messages = [
{"role": "user", "content": "Who are you?"},
]
inputs = tokenizer.apply_chat_template(
messages,
add_generation_prompt=True,
tokenize=True,
return_dict=True,
return_tensors="pt",
).to(model.device)
outputs = model.generate(**inputs, max_new_tokens=40)
print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:]))ABEJA-Qwen3-14B-Agentic-256k-v0.1
ABEJA-Qwen3-14B-Agentic-256k-v0.1は、Alibaba社の開発したQwen3-14Bに対して追加学習を行ったものです。
エージェントとして活用できるようにロングコンテキスト性能とPlanning / Tool Use などの Agentic な能力の向上を狙ったもので、コンテキスト長として256kまで対応しつつ、思考とツール利用のループが出来ることが主な特徴のモデルです。
※本モデルは、幅広い汎用用途向けというより、エージェント利用を主な想定としております。
モデル詳細・使い方
モデルの特徴やサンプルコードは下記ブログを参照してください。
https://tech-blog.abeja.asia/entry/geniac3-qwen3-agentic-model
ベースモデルであるQwen3-14Bと同様に、パラメータとしては、Temperature=0.6, TopP=0.95, TopK=20, MinP=0 及び greedy decodingを使わないことを推奨しています。
For thinking mode, use Temperature=0.6, TopP=0.95, TopK=20, and MinP=0 (the default setting in generation_config.json). DO NOT use greedy decoding, as it can lead to performance degradation and endless repetitions. For more detailed guidance, please refer to the Best Practices section.
開発プロセス・ノウハウ
強化学習を中心とした開発プロセスの詳細及びそこで得られたノウハウについてもこちらで公開しています。
https://tech-blog.abeja.asia/entry/geniac3-agentic-rl-process
開発者
- Fumitaka Iwaki
- Keisuke Fujimoto
- Kyo Hattori
- Shinya Otani
- Tomoki Fujihara
- Yudai Kato
(*)アルファベット順
License
This model is licensed under the Apache License 2.0.
See:
- Downloads last month
- 416

# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="abeja/ABEJA-Qwen3-14B-Agentic-256k-v0.1") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)