K-EXAONE 2.0

Released
2026-07-29
Last refreshed
2026-11-09
Status
Researched 46d ago
Open sourceCommercial use: permittedRAGAgentsLong contextJSON / Tool use

K-EXAONE 2.0 is available now for rag, agents, and long context with open-source and 256K context; evaluate it while provider pricing coverage matures.

LG Research releases · 3 in the last 12 monthsChangelog →
Specifications
Family
K-EXAONE
Released
2026-07-29
Context
256K
Parameters
750B (37B activated)
Architecture
Mixture of Experts
Specialization
general
Openness
Open source
License
Apache 2.0OSI-approvedPermitted
Weights
Available
Code
Unknown
Training
Multi-stage
Created by

Advancing AI for a Better Life

South Korea
Founded 2020
Website
Pricing

No tracked provider pricing is available yet.

About

LG AI Research K-EXAONE 2.0 (HF LGAI-EXAONE/K-EXAONE-2.0-750B-A37B) is an open-weight mixture-of-experts LLM with 750B total and 37B active parameters (256 experts + 1 shared, 8 active; 78 layers with sliding-window/global hybrid attention and an MTP layer), 262,144-token context, reasoning and tool calling, covering 10 languages (Korean, English, Spanish, German, Japanese, Vietnamese, French, Italian, Polish, Portuguese). Unlike K-EXAONE (236B, K-EXAONE license), 2.0 is released under Apache 2.0, in BF16, FP8, NVFP4 and DSpark variants. Technical report: arXiv 2608.04505.

Providers

No active provider routes are tracked for this model yet.

Capabilities

ReasoningJSON / Tool use

Benchmark peer barsfor RAG

No task-mapped benchmark peers are available for this model yet.

Migration checks

No linked migration route is available for this model yet.

API versions

LGAI-EXAONE/K-EXAONE-2.0-750B-A37BK-EXAONE 2.0