GPT-OSS-ZhTW-Thinking

Model on HuggingFace License

A specialized language model optimized for thinking in Traditional Chinese (Taiwanese Mandarin).

🌟 Key Features

  • Native Taiwanese Mandarin Thinking: Default reasoning and thinking patterns optimized for Traditional Chinese
  • Enhanced Cultural Understanding: Deep comprehension of Taiwanese cultural contexts, idioms, and social nuances
  • GPT-based Architecture: Standard GPT-OSS transformer architecture fine-tuned for zh-TW applications

📊 Model Specifications

  • Model Size: 120B parameters
  • Architecture: GPT-based MoE transformer
  • Training: Fine-tuned for Traditional Chinese (zh-TW)

🚀 Usage

Serving with vllm or sglang.

📝 License

This model is released under the Apache 2.0 License.

🤝 Contributing

We welcome contributions and feedback! Please open an issue or submit a pull request if you have suggestions for improvements.


Made with ❤️ by FreeSEED-AI

Downloads last month
-
Safetensors
Model size
117B params
Tensor type
BF16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for FreeSEED-AI/gpt-oss-120b-mandarin-thinking

Finetuned
(38)
this model
Finetunes
1 model
Quantizations
1 model