Inference
EN

DeepSeek V4 Pro 0813 vs GPT-6 Luna

Change models
DeepSeek V4 Pro 0813DeepSeek
GPT-6 LunaOpenAI
IndexAll categories, out of 100.
65.5 ±3.7
65.5 ±2.1
GapIndex points below the top.
Top
Top
CoverageIndex weight that has results.
85%
100% best
SpeedOutput tokens per second.
42/s
66/s best
EloLMArena rating and rank.
—
1388
ReleasedMonth of release.
2026-08
2026-09
Input / 1MUS dollars per 1M input tokens.
$0.111
$0.10 best
Output / 1MUS dollars per 1M output tokens.
$4.44
$0.50 best
Batch output / 1MBatch price per 1M output tokens.
$1.98
$0.25 best
1K chats2K tokens in, 500 out, each.
$2.44
$0.45 best
1K documents20K tokens in, 1K out, each.
$6.66
$2.50 best
1K agent tasks200K tokens in, 20K out, each.
$111.00
$30.00 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
63.7
66.9 best
CodingCode writing and repair.
65.1
66.8 best
ReasoningLogic problems and puzzles.
68.5 best
64.3
MultimodalTasks with images and text.
—
65.0
KnowledgeFacts and expert knowledge.
65.0 best
63.5
MultilingualTasks in many languages.
—
53.3
InstructionTasks with strict prompt rules.
62.0 best
39.8
MathMath problems.
64.1
68.5 best

Benchmarks

43 shared
Only DeepSeek V4 Pro 0813: 44Only GPT-6 Luna: 58
DeepSeek V4 Pro 0813
GPT-6 Luna
Artificial Analysis Humanity's Last ExamKnowledge
41.0% best43 / 222
38.5%54 / 222
Artificial Analysis Long Context ReasoningReasoning
80.3%40 / 217
83.3% best13 / 217
GPQA diamond · Epoch AIKnowledge
91.7% best30 / 217
90.5%39 / 217
Critical Physics TasksReasoning
18.0%35 / 214
19.4% best31 / 214
OTIS Mock AIME 2024-2025Math
98.6%27 / 200
98.9% best22 / 200
Artificial Analysis Omniscience AccuracyKnowledge
49.1% best22 / 164
39.6%54 / 164
GDPval-AA normalizedAgentic
54.5% best23 / 133
46.6%36 / 133
ARC-AGI-2 (semi-private)Reasoning
61.3% best38 / 131
59.3%44 / 131
ARC-AGI-1 (semi-private)Reasoning
90.0% best36 / 126
86.7%46 / 126
Vibe Code Bench v1.1Coding
82.3% best25 / 123
81.6%26 / 123
ARC-AGI-2 (public eval)Reasoning
59.4%42 / 122
61.8% best36 / 122
ARC-AGI-1 (public eval)Reasoning
93.3% best39 / 116
92.5%40 / 116
Artificial Analysis SciCodeCoding
51.0%48 / 114
54.6% best32 / 114
MedScribeHealthcare
80.2%59 / 109
83.7% best42 / 109
MedCodeHealthcare
42.5%50 / 106
44.7% best40 / 106
SimpleQA VerifiedKnowledge
52.9% best26 / 91
41.4%49 / 91
Chess PuzzlesReasoning
46.0% best26 / 90
36.0%42 / 90
FrontierMath-Tiers-1-3-v2-PrivateMath
64.6%39 / 90
78.9% best21 / 90
Legal Research BenchLegal
40.9% best33 / 90
30.3%54 / 90
Tax Agent BenchFinance
26.8% best40 / 82
19.9%55 / 82