Basis AI released Conversations 1500 on September 23, 2026, saying it provides 1,502 hours of multilingual, multi-party conversation for research and commercial use. It is the company's first major dataset update since the release of its earlier conversational data in 2024.
Basis AI reported 1,502 hours of conversation data, measured across 2,396 segments and 2,645 unique speakers. That compares with the earlier dataset, which contained 1,200 hours of conversation.
Conversations 1500 is built on a multilingual, multi-party conversation framework and targets the study of human interaction and the development of better AI models for understanding nuances, turn-taking, and emotional understanding. Availability begins with the full release, initially for researchers and developers worldwide.
"We're making 1,500 hours of conversational data available so that researchers and builders worldwide can study human interactions and develop better models," said Alexis Sursock, Basis AI's community manager. The dataset aims to improve the understanding of human conversations and emotional context.
The announcement follows Basis AI's previous dataset releases and aligns with the company's mission to provide robust data for AI research.
Basis AI did not say when the next dataset will be released, and it raised the question of how to handle the uneven representation of languages in the dataset.
The company plans to continue releasing interaction data to cover the natural modes of human interaction.
Source: huggingface
