Ollama started as a fast way to run open models locally. In 2025 it shifted from a model runner to…