NorwAI to release a ‘free to use’ GPT-model

NorwAI will introduce an open Norwegian-based language model for public use this winter. The model will be available for commercial actorss and public sectors to use for experiments, testing and development of applications. 

- We will use the next coming months to build a new model for people to play with. We hope to finish this job before the end of the year, says Jon Espen Ingvaldsen, adjunct associate professor at NorwAI and project advisor at the Norwegian Large Language Model project at NorwAI. 

Smaller

Consequently, the new open model will be smaller in size than that of the research model NorGPT-23 model NorwAI already has in place. The NorGPT 23 contain for example proprietary news material from Schibsted company released for research purposes.  

- We have not decided yet whether to use the GPT or the Llama architecture. At NorwAI we have used both structures successfully before, says Jon Espen Ingvaldsen.

Immense interest

Nowadays, since the explosion of interest for language models, NorwAI receives lots of questions from people, companies and institutions who want to use the Norwegian language models in their work. NorwAI’s policy is openness in our research work. The coming open model is a response to the immense interest in this field. 

The smaller new model will not affect the continuous work at NorwAI to build a large, national foundation NorGPT-model to serve the Norwegian community for language model service. A research ‘proof-of-concept’ model with 40 billion parameters is planned to be built next year.