What AI should I use?
See which AI models Novelcrafter users use the most
Many of our users find the AI model that they chose for prose generation or in Workshop Chat varies. Different tasks require different AI models to perform them.
Which AI model that you chose may change based on your:
- Budget - the price of AI models can vary from fractions of a cent/million tokens to >$10/million tokens. You can also run open source models locally (on your own device) for free.
- Preferences in output style - model have their own ‘personalities’ and you might prefer some prose/responses over others.
- Context Requirements - some models may be fine for short stories, whereas others are needed for when you’re writing epics.
Which models are used the most?
Finding the LLM that works best for you is easier said than done. That said, we want to help you as much as we can. Below is a list of the most popular AI models in the last month (as of Mid August 2025) in Novelcrafter.
This is not a comprehensive list - we offer more AI models, in both our system model collections and when you create your own model collection but these are the top models used on our platform:
- Claude 3.5, 3.7, and 4 Sonnet
- GPT-4o and GPT-4o mini, with the GPT-5 family growing in popularity
- Deepseek v3 and Deepseek R1
- WizardLM 2 8x22B
- Gemini 2.5 Pro and Flash
- Mistral models (including the new Mistral Medium 3.1)
Which models are best for writing?
This can depend on your writing style and general preferences, however all of the system models included in the default model collections have been curated by us to include options for each price point. As a general rule of thumb:
- For large context windows and deep analysis, choose a Gemini model.
- For natural sounding prose, that can mimic your style, use an Anthropic Claude model such as Sonnet. GPT-5 (not GPT-5 Chat) is also showing promise in mimicking prose styles.
- For NSFW writing, choose an unmoderated/less moderated model, such as Grok, Qwen, Mistral or Deepseek.