Inference
EN

DeepSeek V3.1 (Reasoning) vs K-Exaone

Change models
DeepSeek V3.1 (Reasoning)DeepSeek
K-ExaoneLG AI Research
IndexAll categories, out of 100.
43.0 ±6.4 best
42.3 ±7.9
GapIndex points below the top.
Top
−0.7 · no clear gap
CoverageIndex weight that has results.
85% best
75%
SpeedOutput tokens per second.
—
—
EloLMArena rating and rank.
—
—
ReleasedMonth of release.
2025-08
2025-12
Input / 1MUS dollars per 1M input tokens.
—
—
Output / 1MUS dollars per 1M output tokens.
—
—
1K chats2K tokens in, 500 out, each.
—
—
1K documents20K tokens in, 1K out, each.
—
—
1K agent tasks200K tokens in, 20K out, each.
—
—
ContextMaximum tokens in one request.
128K
256K best
AgenticMulti-step tasks with tools.
27.6
43.1 best
CodingCode writing and repair.
51.4 best
37.7
ReasoningLogic problems and puzzles.
43.5
44.2 best
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
47.0 best
39.6
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
27.8
49.5 best
MathMath problems.
46.5
—

Benchmarks

8 shared
Only DeepSeek V3.1 (Reasoning): 7Only K-Exaone: 2
DeepSeek V3.1 (Reasoning)
K-Exaone
Artificial Analysis Humanity's Last ExamKnowledge
14.3% best128 / 222
13.9%129 / 222
Artificial Analysis Long Context ReasoningReasoning
56.7%144 / 217
61.3% best140 / 217
Critical Physics TasksReasoning
2.0% best105 / 214
1.1%118 / 214
Artificial Analysis GPQA DiamondKnowledge
77.9%106 / 199
78.3% best104 / 199
Artificial Analysis Omniscience AccuracyKnowledge
29.0% best82 / 164
16.4%146 / 164
τ²-Bench Tool-Agent-User EvaluationAgentic
37.4%105 / 149
74.3% best76 / 149
Artificial Analysis IFBenchInstruction
41.5%98 / 146
64.7% best56 / 146
Artificial Analysis AIME 2025Math
89.7%
90.3%