K-EXAONE 2.0 技術報告

摘要

本技術報告介紹 K-EXAONE 2.0，這是由 LG AI Research 開發的開放權重多語言基礎模型，是我們朝向全球前沿規模基礎模型目標邁進的一步。我們並非從零開始訓練，而是對 K-EXAONE 進行升級再利用並擴展其架構，產生了一個總參數量為 750B 的混合專家（MoE）模型，每個 token 激活約 37B 參數——容量是前代模型的三倍以上。K-EXAONE 2.0 支援高達 256K tokens 的上下文長度，並將多語言涵蓋範圍從六種語言擴展至十種。其訓練流程結合持續預訓練、難度導向的中期訓練與後期訓練，以強化推理能力、智能體編程、多語言能力，以及根植於韓國社會文化脈絡的安全性。在九個為反映實際使用條件而選定的評估類別中，K-EXAONE 2.0 相較於 K-EXAONE 有所提升，並與其他開放權重模型保持競爭力，其中在智能體編程與長上下文理解方面進步最為顯著，在長上下文檢索與安全性方面則展現最明確的優勢。K-EXAONE 2.0 以 Apache 2.0 授權條款釋出，使更廣泛的 AI 生態系統得以評估、部署、改編並在此基礎上持續開發，同時也標誌著我們邁向全球前沿挑戰的起點——而非終點。

English

This technical report presents K-EXAONE 2.0, an open-weight multilingual foundation model developed by LG AI Research as a step in our effort toward global frontier-scale foundation models. Rather than training from scratch, we upcycle K-EXAONE and expand its architecture, yielding a Mixture-of-Experts (MoE) model with 750B total parameters and approximately 37B activated per token---more than three times the capacity of its predecessor. K-EXAONE 2.0 supports context lengths of up to 256K tokens and expands multilingual coverage from six to ten languages. Its training pipeline combines continual pre-training, difficulty-focused mid-training, and post-training to strengthen reasoning, agentic coding, multilingual capability, and safety grounded in Korean sociocultural contexts. Across nine evaluation categories selected to reflect the conditions of practical use, K-EXAONE 2.0 improves over K-EXAONE and remains competitive with open-weight models, showing its largest gains in agentic coding and long-context understanding and its clearest strengths in long-context retrieval and safety. Released under the Apache 2.0 license, K-EXAONE 2.0 enables the wider AI ecosystem to evaluate, deploy, adapt, and build upon it, while marking the beginning---rather than the endpoint---of our challenge toward the global frontier.