Inference
EN

DeepSeek-R1 vs DeepSeek V3

Change models
DeepSeek-R1DeepSeek
DeepSeek V3DeepSeek
IndexAll categories, out of 100.
38.9 ±5.0 best
29.9 ±4.9
GapIndex points below the top.
Top
−9.0
CoverageIndex weight that has results.
85%
85%
SpeedOutput tokens per second.
18/s
37/s best
EloLMArena rating and rank.
1373 best
1333
ReleasedMonth of release.
2025-01
2024-12
Input / 1MUS dollars per 1M input tokens.
$0.70
$0.27 best
Output / 1MUS dollars per 1M output tokens.
$2.50
$1.10 best
1K chats2K tokens in, 500 out, each.
$2.65
$1.09 best
1K documents20K tokens in, 1K out, each.
$16.50
$6.50 best
1K agent tasks200K tokens in, 20K out, each.
$190.00
$76.00 best
ContextMaximum tokens in one request.
64K
128K best
AgenticMulti-step tasks with tools.
27.6 best
26.0
CodingCode writing and repair.
42.6 best
31.1
ReasoningLogic problems and puzzles.
39.8 best
34.1
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
43.5 best
34.3
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
28.5
29.1 best
MathMath problems.
43.8 best
27.2

Benchmarks

25 shared
Only DeepSeek-R1: 9Only DeepSeek V3: 9
DeepSeek-R1
DeepSeek V3
Artificial Analysis Humanity's Last ExamKnowledge
15.8% best123 / 222
2.9%220 / 222
Artificial Analysis Long Context ReasoningReasoning
55.7% best146 / 217
29.3%183 / 217
GPQA diamond · Epoch AIKnowledge
71.7% best111 / 217
56.5%131 / 217
Critical Physics TasksReasoning
1.4% best113 / 214
0.0%150 / 214
OTIS Mock AIME 2024-2025Math
53.3% best118 / 200
15.8%150 / 200
Artificial Analysis GPQA DiamondKnowledge
81.3% best94 / 199
55.7%168 / 199
Artificial Analysis Omniscience AccuracyKnowledge
30.5% best78 / 164
25.5%103 / 164
LegalBenchLegal
67.3%133 / 149
80.8% best83 / 149
τ²-Bench Tool-Agent-User EvaluationAgentic
36.5% best107 / 149
22.8%126 / 149
Artificial Analysis IFBenchInstruction
39.6% best105 / 146
34.8%122 / 146
TaxEval v2Finance
72.3% best66 / 143
67.9%104 / 143
Graduate-Level Google-Proof Q&AKnowledge
71.5% best87 / 141
59.1%112 / 141
MMLU ProKnowledge
83.2% best73 / 135
73.8%113 / 135
CorpFin v2Finance
54.1% best97 / 131
52.5%101 / 131
Software Engineering Benchmark VerifiedCoding
49.2% best99 / 114
42.0%107 / 114
AIMEMath
74.0% best52 / 95
27.5%74 / 95
MedQAHealthcare
90.8% best44 / 95
80.9%71 / 95
MATH level 5Math
93.1% best13 / 88
64.9%37 / 88
Massive Multitask Language Understanding ProfessionalKnowledge
84.0% best27 / 75
75.9%48 / 75
MGSMMultilingual
92.3% best20 / 74
92.1%21 / 74