I know, I often do that, but it's still not enough. E.g. things like SmolLM3 which required some llama ccp tweaks wouldn't work via guff for the first week after it had been released.

Just checked: https://github.com/ollama/ollama/issues/11340 still open issue.