Jina AI launches the world’s first open source 8K text embedding model, claiming to surpass OpenAI

IT Home News on October 26, Jina AI issued a press release on its official website, announcing the launch of the jina-embeddings-v2 model, which is currently the only open source product that supports 8K (8192 tokens) context length. Similar in performance to OpenAI's text-embedding-ada-002.

Regarding the MTEB rankings, IT House found that the official explanation was as follows:

Compared with OpenAI's 8K model text-embedding-ada-002, jina-embedding-v2 is better than OpenAI's text-embedding-v2 in terms of classification average, rearrangement average, retrieval average and summary average. ada-002

▲ Picture source Jina AI official website

▲ Picture source Jina AI official website

Officially stated that the jina-embeddings-v2 model was built from scratch. Over the past three months, the Jina AI team has conducted intensive research and development, data collection, and adjustments.

Jina AI also claims that the context length of jina-embeddings-v2 8K is beneficial to the following application scenarios:

• Legal Document Analysis: Ensure every detail in large amounts of legal text is captured and analyzed.
• Medical Research: Comprehensive embedding of scientific papers for advanced analysis and discovery.
• Literary Analysis: Delve into long-form content to capture subtle thematic elements.
• Financial Forecasting: Gain superior insight with detailed financial reporting.
• Conversational AI: Improve chatbot responsiveness to complex user queries.

The above is the detailed content of Jina AI launches the world’s first open source 8K text embedding model, claiming to surpass OpenAI. For more information, please follow other related articles on the PHP Chinese website!