bratao/Qwen3OIE-0.6B

Hugging Face
TEXT GENERATIONPricing:Input $0.04 / Cached $0.008 / Output $0.08Concurrent Unit Cost:1Model Size:0.8BQuant:BF16Context Size:32kTool Calling:SupportedPublished:May 31, 2025License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Warm

bratao/Qwen3OIE-0.6B is a 0.8 billion parameter decoder-only causal language model fine-tuned from Qwen/Qwen3-0.6B. Developed by bratao, this model specializes in abstractive Open Information Extraction (OpenIE) for Portuguese, generating binary extractions with ARG0, V, and ARG1 fields in JSON format from a given sentence. It is the smallest published Qwen3OIE checkpoint, designed for local experimentation and applications requiring structured information extraction from Portuguese text.

Loading preview...

Qwen3OIE-0.6B: Portuguese Abstractive Open Information Extraction

This model, bratao/Qwen3OIE-0.6B, is a 0.8 billion parameter decoder-only causal language model derived from Qwen/Qwen3-0.6B. Its primary function is abstractive Open Information Extraction (OpenIE) for the Portuguese language. Given a single sentence, it generates one or more binary extractions, formatted as JSON, containing ARG0, V, and ARG1 fields.

Key Capabilities & Features

  • Portuguese OpenIE: Specifically fine-tuned for extracting structured information from Portuguese text.
  • Abstractive Generation: The model can normalize or infer wording, rather than strictly copying spans from the input, offering flexibility in output.
  • JSON Output: Extractions are provided in a clear JSON format, making them easy to parse and integrate into applications.
  • Compact Size: As the smallest Qwen3OIE checkpoint, it's suitable for local experiments and resource-constrained environments.
  • Integration: Designed for use with the portuguese-openie library, simplifying deployment and usage.

Use Cases & Considerations

This model is ideal for applications requiring automated extraction of subject-verb-object (or similar) relationships from Portuguese sentences. It's particularly useful for tasks like knowledge graph construction, semantic search, or data enrichment where structured data is needed from unstructured text. Users should be aware that due to its abstractive nature, strict provenance might require validation against the original input. Evaluation on a silver-standard dataset showed a lexical match F1 score of 0.3005. It performs best on encyclopedic Portuguese sentences and may have limitations with highly conversational or specialized text.

Finetunes
1 models