TurkuNLPFinGPT

FinGPT XL

FinGPT XL is a Finnish text-generation base model in TurkuNLP's Finnish GPT-3 series. It uses the BLOOM architecture and is not an OpenAI GPT-3 checkpoint.

Model design and training

24 layers, width 2,064 and 24 attention heads. The card reports 300 billion training tokens. The Finnish corpus combines web text, news, books, Wikipedia and discussion data. This is a pure language model, without dialogue or question-answering instruction tuning.gpt3-finnish-xl: official model card

Practical assessment

Practical guidance. XL crosses into the billion-parameter range without adding layers over Large. Estimate memory from the actual precision and runtime rather than the size label. For task adaptation, retain a baseline on general Finnish so improvements do not hide loss of broader language behavior.

Specifications & access

FinGPT XL specifications
DetailValue and conditions
AccessDownloadable
Model identifiersgpt3-finnish-xlFinnish GPT-3 XL
Model size1.5 billion parameters.gpt3-finnish-xl: official model card
LicenseApache 2.0.gpt3-finnish-xl: official model card
Training statePretrained base; not instruction-tuned.gpt3-finnish-xl: official model card

Related models

  • FinGPT 13B is a Finnish text-generation base model in TurkuNLP's Finnish GPT-3 series.

    Launch date not confirmedDownloadable
  • FinGPT Small is a Finnish text-generation base model in TurkuNLP's Finnish GPT-3 series.

    Launch date not confirmedDownloadable