Ollama · Server and deployment

Keep a model loaded longer

Keep models in memory for an hour after the last request instead of the default five minutes.