Anthropic’s ChatGPT competitor can now analyze hundreds of pages at once



summary
Summary

Anthropic’s ChatGPT competitor Claude can now handle hundreds of pages of text, overtaking GPT-4.

The context window of a chatbot like ChatGPT determines the amount of text the AI ​​model can process at a time. In OpenAI’s currently available GPT-4, the window is just under 8,000 tokens, and a variant with just under 32,000 tokens is in closed testing.

AI startup Anthropic is now expanding the context window of its ChatGPT competitor Claude from just under 9,000 tokens to 100,000 tokens. According to the company, this is the equivalent of about 75,000 words, or hundreds of pages of text.

Anthropic Claude can process hundreds of pages

Claude is said to be able to process that amount in less than a minute. In one example, the company changed a sentence in the classic “The Great Gatsby” and asked Claude if anything had changed. Claude found the changed sentence in 22 seconds.

ad

With the large context window, Claude should be able to combine knowledge from multiple documents or an entire book for businesses, for example. ”For complex questions, this is likely to work substantially better than vector search based approaches,” says Anthropic.

Companies are already using vector databases from vendors such as Pinecone to store large amounts of text and make it processable for ChatGPT or other language models. Whether the context window actually offers advantages over such vector databases remains to be seen.

Anthropic sees numerous applications that could benefit from the context window

Claude could create summaries of transcribed podcasts – 100,000 tokens equals about six hours of audio – or make hundreds of pages of API documentation searchable, according to Anthropic.

Recommendation

Access to Claude can be requested here.

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top