WestCode1357/gpt-sw3-1.3b-instruct
The WestCode1357/gpt-sw3-1.3b-instruct is a 1.3 billion parameter instruction-tuned causal language model, a community mirror of AI Sweden's GPT-SW3. It is optimized for fast inference and good chat quality primarily in Swedish, also supporting Norwegian, Danish, Icelandic, and English. This model is intended for research and educational purposes, excelling in multilingual Nordic language generation within its 2048 token context length.
Loading preview...
GPT-SW3-1.3b-instruct: A Multilingual Nordic Chat Model
This model, gpt-sw3-1.3b-instruct, is a 1.3 billion parameter instruction-tuned language model, serving as a community mirror of AI Sweden's original GPT-SW3. It is designed for fast inference and strong chat capabilities, particularly in Swedish, while also supporting Norwegian, Danish, Icelandic, and English. The model was developed by AI Sweden in collaboration with RISE and WASP WARA for Media and Language, trained on 320 billion tokens across these languages and code.
Key Capabilities
- Multilingual Generation: Proficient in Swedish, Norwegian, Danish, Icelandic, and English.
- Instruction Following: Fine-tuned for chat and instruct-based interactions.
- Efficient Inference: Optimized for speed, making it suitable for applications requiring quick responses.
- Research Focus: Primarily intended for scientific and research use, reflecting biases from its training data.
Intended Use
This model is provided for research and educational purposes only. Due to potential biases from its large-scale web training data and lack of extensive safety alignment, it is not intended for commercial use or deployment in consumer-facing products without significant additional evaluation and safety measures. Users are responsible for content generated and should review the AI Sweden RAIL license before any production deployment.