New Upload Categories: Children’s Books by Age Range
Insight now includes targeted document analysis for children’s books in the following age ranges and categories:
- 0–5 fiction
- 0–5 nonfiction
- 6–8 fiction
- 6–8 nonfiction
- 8–10 fiction
- 8–10 nonfiction

For 0–5 and 6–8 age range categories, Insight automatically extracts and describes images because picture books convey more of the story through images.
Try it: In the upload screen, select “Children’s Book” in the document type selector. An “Age range” selector will appear, from which you can select a sub-category.
Thema Upgrade
We have implemented Thema 1.6 and optimized the AI process for identifying categories for your titles.
Try it: This process will be used on all newly uploaded books and documents. For titles already in Insight, click the Edit (pencil) icon next to Thema and then click the “Regenerate” button.
Upgraded Anthropic LLMs in AI Chat
Following Anthropic’s release of updated LLMs, Insight’s AI Chat now includes the latest version of Haiku (3.5), and we have removed Claude Opus.
- Haiku’s improved reasoning and content understanding, combined with its ability to process large amounts of data quickly, makes it a strong option for interacting with long texts.
- With upgraded capabilities, Sonnet 3.5 and even the smaller Haiku models now surpass Opus in reasoning, content understanding, and the quality of its chat responses. For Insight users who regularly interact with Opus, we recommend switching to Sonnet 3.5.
Try it: Open AI Chat and test your favorite prompts with Claude Sonnet 3.5 and Claude Haiku.
Optimized Language Prompts
We’ve updated the way we handle multiple languages within Insight’s automatic analysis. The result is more consistent use of your selected output language.
Improved Document Summarization Process
Summarization is one of three ways—along with vectorization and storing the full-length text— that Insight processes your long documents and books. We’ve optimized our summarization process so that Insight generates more accurate results and responses.
