Local Models (Ollama & LM Studio)

Novel Mage supports fully local AI models through Ollama and LM Studio the two most widely used tools for running open-source LLMs on your own hardware. When you use a local model, every part of your writing workflow is offline: the AI runs on your CPU or GPU, your manuscript never leaves your machine, and there are no API costs, no token limits, and no internet dependency.

For writers, local AI is about more than privacy though that matters enormously. It's about uncapped usage. When the AI runs locally, you can generate, rewrite, summarize, and chat with no credit counter ticking, no usage tier limiting your session length, and no monthly bill for the AI component. For heavy users, this changes the economics of AI-assisted writing entirely. A local 13B model running on a consumer GPU has zero marginal cost per generation.

Novel Mage lets you mix local and cloud AI in the same project. You might use a fast local model for first-draft generation where raw speed matters, then switch to Claude for nuanced rewriting where quality is the priority all within the same novel, per task. This hybrid approach gives you the cost efficiency of local AI plus the output quality of frontier models where it counts. Local model support works with any Ollama-compatible or LM Studio-compatible model, including Llama 3, Mistral, Qwen, DeepSeek, and dozens of fine-tuned variants.