South Korea236B / 23B active256K contextReasoning

Explore K-EXAONE 236B-A23B

Use LG AI Research's multilingual mixture-of-experts flagship for Korean reasoning, long-context analysis, coding, and agentic tool use.

Model

LGAI-EXAONE/K-EXAONE-236B-A23B

Input

Configure your request

System prompt Optional

Set the assistant role, language, or response policy.

Conversation history Optional

Optional previous system, user, assistant, and tool messages as JSON.

Your message

Parameters

Thinking modeUse deep reasoning; enabled by default for K-EXAONE. Stream response Sample output

Temperature

LG’s recommended setting.

Top P

LG’s recommended setting.

Top K

Maximum output tokens

Prompt and output together must fit the 256K context.

Presence penalty

LG recommends zero.

Frequency penalty

Repetition penalty

Seed

Stop sequences

Function tools (JSON)

Supports OpenAI and Hugging Face tool schemas.

Tool choice