Apertus Model

[hammad93]

This issue tracks experiments and deployment of the Apertus model. This is a fully open, transparent, multilingual language model. As of this writing, the quantized version is available through Hugging Face from the devquasar organization but others maybe considered. Please reference the orchestration branch.

[hammad93]

sudo huggingface-cli download DevQuasar/swiss-ai.Apertus-70B-Instruct-2509-GGUF --include "*Q8*" --local-dir .

[hammad93]

Apertus 70b-Instruct quantized at 8 bits is successfully integrated into the workflow.

We have a pull request here, #4 pending conditions to main branch.