Insight is build to harness the best of the leading AI models available. The platform uses different models in the background to produce the automatic analysis of your books and long documents.
In addition, multiple LLMs are available to use in AI Chat, so you can get to know how various models perform and which work best for your tasks. (Take a look at our favorite chat prompts for publishers.)
How to Choose an LLM
To get started: You can’t go wrong with either GPT 4o or Sonnet 3.5. Both are among the most advanced available LLMs from their respective developers—OpenAI and Anthropic. Gemini Pro is another top choice. You’ll find variations in how these models interpret your prompts and structure their responses.
Compare results: If you are working on a prompt and not quite getting the results you want, try using the same prompt with GPT 4o, Sonnet 3.5, and Gemini Pro.
Keep testing: As you work more with AI Chat, you may prefer the results of one model over another. However, these LLMs are regularly updated and their capabilities change. Now and then, test your favorite prompts across multiple LLMs to see if one is generating improved results over the others.
LLMs in Insight AI Chat
GPT 4o — This is the leading model from OpenAI and one of the most well-known. It reliably excels at general tasks, reasoning, and producing natural-sounding text. It’s a benchmark model against which the others are often compared, and is one of Insight’s favorite models for most jobs.
GPT mini — This is a smaller model from OpenAI, which can process more information quickly. This model can process and generate longer texts than GPT 4o, but its reasoning is not as sophisticated.
Claude Sonnet 3.5 — This leading model from Anthropic is nearly on par with GPT 4o. It excels in reasoning, processing information, and producing quality responses, and it’s another of our top LLMs for daily tasks.
Claude Haiku — This small model from Anthropic has been recently updated (we’re using 3.5 Haiku) and is among the leading small LLMs that can process and understand large amounts of information quickly and accurately. Haiku produces streamlined answers and text.
Gemini Pro — From Google, Gemini Pro is on par with GPT 4o and Sonnet 3.5 for many tasks. You may find that Gemini Pro’s answers are structured in a slightly different way compared to the other models. This is a great third option when you’re looking for variations in AI Chat answers.
Mistral — Developed by a France-based company, Mistral Large is available in Insight’s AI Chat. This model includes advanced multilingual capabilities, and Mistral’s various models are quickly catching up to the leading AI providers.
DALL-E 3 — Image generator from Open AI
Imagen 2 — Image generator from Google