Mi:dm 2.0

Mi:dm 2.0 is KT's bilingual large language model built specifically to advance Korea-centric AI. Beyond Korean text processing, it is designed around Korean cultural context, reasoning patterns, and commonsense knowledge, using a custom Korean-optimized tokenizer, proprietary data cleansing, and curriculum learning over high-quality synthetic data.

The lineup ships in two configurations: Mi:dm 2.0 Base with 11.5B parameters built via depth-up scaling, and Mi:dm 2.0 Mini with 2.3B parameters for resource-constrained environments. The models are released under the MIT license, which is unusually permissive for a telecom-backed sovereign model, and KT reports state-of-the-art results on Korean-specific benchmarks including KMMLU.