Upstage Launches the '1T Club'

Upstage is launching the '1T Club' to solve the shortage of Korean-language data and achieve independence for Korean LLMs through the development of high-performance LLM (Large Language Model) technology. The '1T Club' is composed of partner companies that contribute more than 100 million words of Korean-language data, including text, books, and papers.

Upstage recently proved its technical prowess by ranking No. 1 in the world on Hugging Face's 'Open LLM Leaderboard.' Currently, global big-tech models such as Meta's Llama 2 and Google's LaMDA have an overwhelmingly high proportion of English data, which limits their ability to reflect Korean sentiment and regional information.

The '1T Club' aims to build an ecosystem where data providers and model makers can thrive together, offering partner companies the following benefits:

• API usage fee discount: Depending on the number of tokens contributed, partners can use Upstage's high-performance LLM API for free or at a discounted rate
• Profit Share: A portion of the revenue generated through the LLM API business will be distributed according to the amount of data contributed

For data security, the data provided will be used only for pre-training the model. Upstage plans to protect personal information and fundamentally block the extraction of original text or external leaks by introducing Jailbreak Check technology.