Emaoso/Tangshi

TEXT GENERATIONConcurrency Cost:1Model Size:0.5BQuant:BF16Ctx Length:32kTool Calling:SupportedPublished:Jun 1, 2026License:apache-2.0Architecture:Transformer0.0K Open Weights Cold

Emaoso/Tangshi is a 0.5 billion parameter causal language model, fine-tuned by Emaoso from Qwen2.5-0.5B, specifically optimized for generating classical Chinese poetry. It excels at automatically composing five-character (五言) and seven-character (七言) quatrains (绝句) and regulated verses (律诗). Trained on 57,000 Tang Dynasty poems, this model is designed for specialized classical poetry creation.

Loading preview...

Tangshi: Classical Chinese Poetry Generation Model

Tangshi is a specialized large language model developed by Emaoso, fine-tuned from Qwen2.5-0.5B, with 0.5 billion parameters. Its primary focus is the automatic generation of classical Chinese poetry, particularly Tang Dynasty styles.

Key Capabilities

  • Poetry Generation: Expertly creates five-character (五言) and seven-character (七言) quatrains (绝句) and regulated verses (律诗).
  • Specialized Training: Optimized using a dataset of 57,000 Tang Dynasty poems, ensuring high fidelity to classical poetic forms and styles.
  • Model Availability: Provided in two formats:
    • model.safetensors: Native Transformers weights for Python integration.
    • model-ollama.gguf: GGUF quantized weights for local deployment with Ollama, including a pre-configured Modelfile.

Good For

  • Developers and enthusiasts requiring a dedicated model for generating classical Chinese poetry.
  • Applications focused on cultural heritage, educational tools, or creative writing assistance in the domain of traditional Chinese literature.
  • Users looking for a lightweight yet specialized model for poetic composition.