Inference
EN

DeepSeek V4 Flash 0731 vs DeepSeek V4 Pro 0813

Change models
DeepSeek V4 Flash 0731DeepSeek
DeepSeek V4 Pro 0813DeepSeek
IndexAll categories, out of 100.
61.9 ±3.8
65.5 ±3.7 best
GapIndex points below the top.
−3.6 · no clear gap
Top
CoverageIndex weight that has results.
85%
85%
SpeedOutput tokens per second.
61/s best
42/s
EloLMArena rating and rank.
—
—
ReleasedMonth of release.
2026-07
2026-08
Input / 1MUS dollars per 1M input tokens.
$0.043 best
$0.111
Output / 1MUS dollars per 1M output tokens.
$0.30 best
$4.44
Batch output / 1MBatch price per 1M output tokens.
$0.144 best
$1.98
1K chats2K tokens in, 500 out, each.
$0.237 best
$2.44
1K documents20K tokens in, 1K out, each.
$1.17 best
$6.66
1K agent tasks200K tokens in, 20K out, each.
$14.68 best
$111.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
59.9
63.7 best
CodingCode writing and repair.
62.8
65.1 best
ReasoningLogic problems and puzzles.
65.5
68.5 best
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
58.6
65.0 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
59.2
62.0 best
MathMath problems.
61.5
64.1 best

Benchmarks

78 shared
Only DeepSeek V4 Flash 0731: 7Only DeepSeek V4 Pro 0813: 9
DeepSeek V4 Flash 0731
DeepSeek V4 Pro 0813
Artificial Analysis Humanity's Last ExamKnowledge
38.6%53 / 222
41.0% best43 / 222
Artificial Analysis Long Context ReasoningReasoning
79.7%45 / 217
80.3% best40 / 217
GPQA diamond · Epoch AIKnowledge
91.0%33 / 217
91.7% best30 / 217
Critical Physics TasksReasoning
16.6%46 / 214
18.0% best35 / 214
OTIS Mock AIME 2024-2025Math
94.4%47 / 200
98.6% best27 / 200
Artificial Analysis GPQA DiamondKnowledge
90.8%30 / 199
92.8% best14 / 199
Artificial Analysis Omniscience AccuracyKnowledge
40.4%49 / 164
49.1% best22 / 164
LegalBenchLegal
77.7%107 / 149
82.4% best72 / 149
TaxEval v2Finance
70.7%88 / 143
73.1% best52 / 143
LiveCodeBenchCoding
87.3%16 / 142
87.5% best13 / 142
Graduate-Level Google-Proof Q&AKnowledge
88.1%37 / 141
90.1% best29 / 141
GPQA Diamond · Vals AIKnowledge
89.9%27 / 135
92.4% best16 / 135
MMLU ProKnowledge
86.2%45 / 135
87.0% best35 / 135
GDPval-AA normalizedAgentic
46.7%35 / 133
54.5% best23 / 133
ARC-AGI-2 (semi-private)Reasoning
61.4% best36 / 131
61.3%38 / 131
CorpFin v2Finance
61.8%51 / 131
65.4% best27 / 131
ARC-AGI-1 (semi-private)Reasoning
89.0%39 / 126
90.0% best36 / 126
Vibe Code Bench v1.1Coding
74.7%43 / 123
82.3% best25 / 123
ARC-AGI-2 (public eval)Reasoning
63.9% best33 / 122
59.4%42 / 122
ARC-AGI-1 (public eval)Reasoning
94.8% best35 / 116
93.3%39 / 116