Inference
EN
LG AI Research

K-EXAONE 2.0

K-EXAONE 2.0 is a reasoning model from LG AI Research. 17 benchmarks count toward its index, in 6 categories.

  • Open weights
  • Reasoning
  • Input: text
  • Released 2026-07
RankedEnough results for an index.
IndexAll categories, out of 100.48.8±5.4Median 51.0
CoverageIndex weight that has results.85%
SpeedOutput tokens per second.—Median 67/s
First tokenSeconds to the first output token.—Median 1.3 s
Input / 1MUS dollars per 1M input tokens.—Median $0.60
Output / 1MUS dollars per 1M output tokens.—Median $2.55
ContextMaximum tokens in one request.262KMedian 262K
EloLMArena rating and rank.—Median 1422

Run locally

749.36B parameters
16-bit
1574.7 GB
8-bit
837.0 GB
4-bit
478.0 GB
Weights
Hugging Face (opens in a new tab) ↗

Memory need with a short context. The 4-bit copy needs more than 128 GB. Compare local models

Capabilities

Index per category, out of 100.
Out of 100
AgenticMulti-step tasks with tools.3 counted56.493 / 218
CodingCode writing and repair.2 counted49.6140 / 209
ReasoningLogic problems and puzzles.2 counted42.2152 / 227
MultimodalTasks with images and text.None counted——
KnowledgeFacts and expert knowledge.5 counted43.7173 / 268
MultilingualTasks in many languages.None counted——
InstructionTasks with strict prompt rules.2 counted53.130 / 63
MathMath problems.3 counted51.082 / 168

Results

17 counted
BenchmarkThe benchmark name.ResultThe score from the source.PlacePlace among results on this benchmark.
Claw-EvalAgenticAbove fitFar above this model's index.77.7%3 / 42
Instruction-Following EvalInstruction92.4%18 / 70
MMMLUKnowledge86.6%6 / 15
Massive Multitask Language Understanding ProfessionalKnowledge83.5%30 / 75
Artificial Analysis GPQA DiamondKnowledge82.9%89 / 199
Artificial Analysis Humanity's Last ExamKnowledge18.6%120 / 222
Critical Physics TasksReasoning0.9%124 / 214
GDPval-AA normalizedAgentic15.4%87 / 133
Instruction Following BenchmarkInstruction72.6%38 / 57
Artificial Analysis Long Context ReasoningReasoning56.2%145 / 217
Software Engineering Benchmark VerifiedCoding68.2%77 / 114
IMOAnswerBenchMath78.6%7 / 10
Artificial Analysis SciCodeCoding42.0%80 / 114
AIME 2026Math92.3%24 / 34
Scientific Code BenchmarkCoding37.4%25 / 35
GPQA Diamond · BenchLMKnowledge82.2%62 / 84
Harvard-MIT Mathematics Tournament February 2026Math78.4%22 / 27
Humanity's Last ExamKnowledge18.3%65 / 77
Terminal-Bench 2.1 · BenchLMAgentic43.8%77 / 86
Artificial Analysis Omniscience AccuracyKnowledge14.5%155 / 164
PolyMathMultilingual71.3%—

Sources

BenchLMCC BY-NC 4.0

History

5 releases
Index50.4 → 48.8
23 Oct 202612 Nov 2026

Questions

How good is K-EXAONE 2.0?
Its index is 48.8 out of 100. That is place 156 of 283 ranked models.
What is K-EXAONE 2.0 best at?
Agentic: 56.4 out of 100, place 93 of 218.
What is the context window of K-EXAONE 2.0?
262K tokens.
Is K-EXAONE 2.0 open source?
Yes. Its weights are published.
When was K-EXAONE 2.0 released?
July 2026.

Badge

K-EXAONE 2.0 on Inference360: place 156 of 283, index 48.8
HTML
<a href="https://inference360.ai/models/k-exaone-2-0"><img src="https://inference360.ai/badges/k-exaone-2-0.svg" alt="K-EXAONE 2.0 on Inference360: place 156 of 283, index 48.8" height="20"></a>
Markdown
[![K-EXAONE 2.0 on Inference360: place 156 of 283, index 48.8](https://inference360.ai/badges/k-exaone-2-0.svg)](https://inference360.ai/models/k-exaone-2-0)

Compare with

Spark-X2.6-4B48.7K-Exaone42.3

More from LG AI Research

All models
K-Exaone42.3Exaone 4.0 32B32.4Exaone 4.0 1.2BUnranked