DeepSeek V4 Pro 0813 vs GPT-5.6 Luna

Change models
IndexAll categories, out of 100.
63.9 ±4.6
64.3 ±2.1 best
GapIndex points below the top.
−0.4 · no clear gap
Top
CoverageIndex weight that has results.
85%
100% best
SpeedOutput tokens per second.
93/s best
70/s
EloLMArena rating and rank.
—
1431
ReleasedMonth of release.
2026-08
2026-07
Input / 1MUS dollars per 1M input tokens.
$1.32
$0.20 best
Output / 1MUS dollars per 1M output tokens.
$3.96
$1.20 best
Batch output / 1MBatch price per 1M output tokens.
$1.98
$0.60 best
1K chats2K tokens in, 500 out, each.
$4.62
$1.00 best
1K documents20K tokens in, 1K out, each.
$30.36
$5.20 best
1K agent tasks200K tokens in, 20K out, each.
$343.20
$64.00 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
58.5
63.6 best
CodingCode writing and repair.
65.3
66.2 best
ReasoningLogic problems and puzzles.
67.4 best
65.0
MultimodalTasks with images and text.
—
63.7
KnowledgeFacts and expert knowledge.
61.9 best
61.4
MultilingualTasks in many languages.
—
53.0
InstructionTasks with strict prompt rules.
62.6 best
47.8
MathMath problems.
64.3
67.2 best

Benchmarks

44 shared
Only DeepSeek V4 Pro 0813: 2Only GPT-5.6 Luna: 60
DeepSeek V4 Pro 0813
GPT-5.6 Luna
91.7% best34 / 220
91.6%35 / 220
98.6% best30 / 203
98.3%31 / 203
82.4%72 / 149
84.0% best43 / 149
TaxEval v2Finance
73.1%52 / 143
76.2% best6 / 143
61.3% best42 / 137
59.5%47 / 137
92.4% best16 / 135
91.7%18 / 135
MMLU ProKnowledge
87.0% best35 / 135
86.0%50 / 135
90.0% best39 / 132
88.0%46 / 132
CorpFin v2Finance
65.4% best27 / 131
64.2%38 / 131
82.3% best30 / 130
77.1%45 / 130
59.4%45 / 127
60.2% best41 / 127
93.3% best43 / 116
90.8%47 / 116
MedScribeHealthcare
80.2%61 / 111
84.4% best38 / 111
MedCodeHealthcare
42.5% best51 / 108
42.4%52 / 108
40.9% best38 / 98
36.5%53 / 98
52.9% best29 / 94
41.0%54 / 94
Chess PuzzlesReasoning
46.0% best29 / 93
43.0%34 / 93
64.6%42 / 93
82.1% best21 / 93
26.8% best43 / 89
24.8%49 / 89
SWE-benchCoding
96.4% best2 / 87
93.0%9 / 87