AI NewsWaveKrafton Unveils Preview of Raon-1 mini, a 30B-Class LLM
Key points
- Krafton unveils preview of 30B-class LLM Raon-1 mini
- Entire process from design to training handled in-house
- Released on Hugging Face under the Apache 2.0 license
Krafton unveiled on Oct. 21 a preview version of Raon-1 mini, a large language model (LLM) with 30 billion (30B) parameters that it developed in-house. Chosun Biz reported that Krafton handled the entire process, from model design to training, on its own, and wrote that the model supports Korean and English.
iNews24 reported that the model is a version shown ahead of its official release and has been released on Hugging Face under the Apache 2.0 license. Industry News said the model raises response speed and lowers inference costs.
On its structure, EBN reported that the model reduces computational load by activating only about 13% of its total parameters. EBN said the model strengthens Korean conversation and AI agent capabilities to improve cost efficiency relative to performance, and used the phrase "cost-effective LLM" in its headline.