WestCode1357/gpt-sw3-126m-instruct

TEXT GENERATIONConcurrency Cost:1Model Size:0.2BQuant:BF16Ctx Length:2kPublished:May 19, 2026License:otherArchitecture:Transformer Cold

WestCode1357/gpt-sw3-126m-instruct is a 126 million parameter instruction-tuned causal language model, a community mirror of AI Sweden's GPT-SW3 series. With a 2048 token context length, it is specifically designed for rapid loading and prototyping, making it ideal for initial testing and development. Its primary strength lies in its multilingual capabilities, supporting Swedish, Norwegian, Danish, Icelandic, and English.

Loading preview...

gpt-sw3-126m-instruct Overview

WestCode1357/gpt-sw3-126m-instruct is a compact, instruction-tuned language model with 126 million parameters, serving as a community mirror of the original AI Sweden GPT-SW3 model. This model is engineered for instant loading and rapid prototyping, making it highly suitable for initial development and testing phases where speed and resource efficiency are paramount.

Key Capabilities

  • Multilingual Support: Proficient in Swedish, Norwegian, Danish, Icelandic, and English.
  • Instruction Following: Designed to respond to instructions using a specific chat format (<|endoftext|><s>User: [message]<s>Bot: [response]<s>...).
  • Lightweight: Its small size (126M parameters) ensures quick deployment and low computational overhead.

Good For

  • Rapid Prototyping: Excellent for quickly testing ideas and developing initial concepts due to its fast loading times.
  • Research and Education: Intended for scientific and research use in controlled settings.
  • Nordic Language Applications: Particularly strong for tasks involving Swedish, Norwegian, Danish, and Icelandic.

Important Considerations

This model is provided as-is for research and educational purposes only. It was trained on large-scale web data and may contain biases, potentially generating inaccurate, offensive, or inappropriate content. It has not undergone alignment or safety tuning beyond its original training. It is not intended for commercial use and requires thorough evaluation and additional safety measures before any production deployment. Users are responsible for content generated.

Finetunes
1 models