K-EXAONE 2.0
LG AI Research's 750-billion-parameter foundation model, released under Apache 2.0 for unrestricted commercial use and published on Hugging Face -- the third-phase, largest release of LG's EXAONE/K-EXAONE family, more than triple the scale of the prior 236B-parameter model.
Find alternatives to K-EXAONE 2.0
What it does
- Text generation
- Model inference
- Code generation
- Agentic coding
How it compares
-
Alibaba Cloud Model Studio
It also delivered performance comparable to leading global models, including GLM-5.1, DeepSeek V4 Pro Max and Qwen3.5, in instruction-following benchmarks.
News reporting · 19 Sep 2026 -
DeepSeek API Platform
It also delivered performance comparable to leading global models, including GLM-5.1, DeepSeek V4 Pro Max and Qwen3.5, in instruction-following benchmarks.
News reporting · 19 Sep 2026 -
Z.ai Model API
In global and Korean long-context comprehension benchmarks (OpenAI-MRCR and Ko-LongBench), the model scored 94.4 and 89.6, respectively, outperforming Zhipu's GLM-5.1.
News reporting · 19 Sep 2026
Sources
- Pricing
-
Official documentation · 11 Sep 2026
Direct quote: "transitioned the model's license to Apache 2.0, allowing commercial use without restrictions."
- Description
-
Official documentation · 11 Sep 2026
"750 billion parameters... more than a threefold increase in scale compared to the first-phase model (236B)."
- What it does
-
LG AI Research release: "performance across three major coding and agentic coding benchmarks improved by approximately 30%"; adds code-generation and agentic-coding to the existing text-generation and model-inference.
- Sold within
-
Official documentation · 15 Sep 2026
lgresearch.ai's own news page (already cited official-docs elsewhere on this record); downloaded directly from Hugging Face.
- Name
-
Official documentation · 21 Oct 2026
News page headline: "LG AI Research Unveils K-EXAONE 2.0, Korea's Largest 750-Billion-Parameter AI Foundation Model".
- URL
-
Official documentation · 11 Sep 2026
LG AI Research's own news page for K-EXAONE 2.0.
- Deployment
-
Official documentation · 11 Sep 2026
Announcement frames access entirely as a Hugging Face download; no managed hosted endpoint mentioned.
- Modalities
-
Official documentation · 1 Oct 2026
Page: 'the model now supports French, Italian, Portuguese, and Polish / performance across three major coding and agentic coding benchmarks improved by approximately 30%'.
Also from LG Corp
-
ChatEXAONE AI agent
AI work agent that plans, carries out and refines a user's tasks across documents, images, web content and data, runs code, drafts documents in a Canvas and produces evidence-backed deep-research reports, personalised to the user's role and response style.
-
ChatEXAONE Notebook Application
AI workspace that gathers uploaded files, web search results and URLs into one Notebook, answers questions across the selected sources with links back to the supporting passages, and turns the sources into reports, presentations, data tables and mind maps.
-
EXAONE Data Foundry Application
LLM tuning platform that turns a customer's domain documents into question-and-answer training data, refines the generated data with expert preference signals gathered through structured surveys, and carries the data through validation to model tuning.
Something wrong here? Send a correction — quote this product id: lg-corp-k-exaone.