Its not possible to run the latest model architectures without 'moving fast'. The only thing broken here is that they are trying to use an old version with a new model.
and Ollama suffered the same fate when wanting to try new models
and Ollama suffered the same fate when wanting to try new models