Jina AI
A technology company focused on multimodal AI and neural search, offering a range of AI infrastructure tools including a web content scraping and conversion service (r.jina.ai).
Timeline (last 90 days)
In a commentary article on April 13, 2023, Jina AI's Han Xiao calculated, based on GPT-4 prices at the time ($0.03 per thousand input tokens and $0.06 per thousand output tokens), that Auto-GPT costs about $0.29 per step, and a small task of about 50 steps costs about $14.
Dr. Xiao Han's team at Jina AI proposed a deferred chunking scheme: letting the model first read the entire text in full, and only splitting at natural semantic boundaries before outputting the result.
r.jina.ai is a web content scraping service provided by Jina AI that converts web pages into structured Markdown text for large language models to read.
Late Chunking was proposed by Jina AI; it first processes an entire document through a long-context encoder to obtain context-aware representations for each token, then splits by boundaries into chunks, so that the embedding of each chunk retains the full-document context information.
All Facts (4)
In a commentary article on April 13, 2023, Jina AI's Han Xiao calculated, based on GPT-4 prices at the time ($0.03 per thousand input tokens and $0.06 per thousand output tokens), that Auto-GPT costs about $0.29 per step, and a small task of about 50 steps costs about $14.
50%UnverifiedDr. Xiao Han's team at Jina AI proposed a deferred chunking scheme: letting the model first read the entire text in full, and only splitting at natural semantic boundaries before outputting the result.
50%Unverifiedr.jina.ai is a web content scraping service provided by Jina AI that converts web pages into structured Markdown text for large language models to read.
50%UnverifiedLate Chunking was proposed by Jina AI; it first processes an entire document through a long-context encoder to obtain context-aware representations for each token, then splits by boundaries into chunks, so that the embedding of each chunk retains the full-document context information.
50%