Models

This section gives an overview of the models available through the DFM project. The models are available through the Huggingface model hub. To avoid duplicating information surrounding the models and the information regarding the models are available at the models model sheet.

Model recommendations

Danish foundation models maintains a list of state-of-the-art recommendations This list is updated approximately once per year to reflect the best available models for various tasks in Danish language and speech processing.

Text Models

Model	Model type	Size (parameters)
munin-7b-alpha	Decoder	7.24B
dfm-sentence-encoder-large	Encoder	large (355M)
dfm-sentence-encoder-medium	Encoder	medium (110M)

Previously released models

Previously the DFM project released the following text models, however these models were taken down due to copyright concerns. Preventative measures have been taken to ensure that future models do not have the same issues.

Model	Model type	Size (parameters)
encoder-large-v1	Encoder	large (355M)
encoder-medium-v1	Encoder	medium (110M)
encoder-small-v1	Encoder	small (22M)

Similarly the DFM project previously released the following speech models which were also taken down due to copyright concerns.

Model	Model type
xls-r-300m-danish	Pretrained wav2vec2.0 model
xls-r-300m-danish-nst-cv9	Automatic speech recognition
chcaa/xls-r-300m-nst-cv9-da	Automatic speech recognition

We refer to our state-of-the-art model recommendations for an best alternatives to these models.