LG AI Research has released K-EXAONE 2.0, a large-scale multilingual language model that is competitive with leading open-weight models. The model was significantly scaled up from its predecessor through a process of upcycling, continual pretraining, difficulty-focused mid-training, and post-training. K-EXAONE 2.0 shows notable improvements in long-context retrieval and safety, and includes technical details on mitigating activation explosions in deeper layers. AI
IMPACT This release offers a new large-scale, open-weight multilingual model with competitive performance, particularly in long-context retrieval and safety.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Hugging Face Trending Models →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
