DeepSeek V4 Pro 0423 vs DeepSeek V3.2

Change models
IndexAll categories, out of 100.
57.6 ±4.8 best
44.2 ±4.9
GapIndex points below the top.
Top
−13.4
CoverageIndex weight that has results.
85%
85%
SpeedOutput tokens per second.
89/s best
19/s
EloLMArena rating and rank.
1451 best
1425
ReleasedMonth of release.
2026-04
2025-12
Input / 1MUS dollars per 1M input tokens.
$0.892
$0.26 best
Output / 1MUS dollars per 1M output tokens.
$1.78
$0.38 best
1K chats2K tokens in, 500 out, each.
$2.68
$0.71 best
1K documents20K tokens in, 1K out, each.
$19.63
$5.58 best
1K agent tasks200K tokens in, 20K out, each.
$214.15
$59.60 best
ContextMaximum tokens in one request.
1.05M best
164K
AgenticMulti-step tasks with tools.
55.9 best
43.1
CodingCode writing and repair.
59.2 best
48.8
ReasoningLogic problems and puzzles.
54.6 best
42.9
MultimodalTasks with images and text.
—
—
KnowledgeFacts and expert knowledge.
59.9 best
43.2
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
52.7 best
34.8
MathMath problems.
57.6 best
46.8

Benchmarks

10 shared
Only DeepSeek V4 Pro 0423: 31Only DeepSeek V3.2: 28
DeepSeek V4 Pro 0423
DeepSeek V3.2
89.6% best44 / 217
77.3%96 / 217
96.7% best36 / 200
68.4%103 / 200
80.3% best87 / 149
66.0%135 / 149
87.5% best13 / 142
69.9%91 / 142
89.4% best30 / 135
76.3%78 / 135
MMLU ProKnowledge
87.2% best32 / 135
83.1%74 / 135
CorpFin v2Finance
61.4% best52 / 131
47.9%110 / 131
56.2% best14 / 66
34.8%42 / 66
IOI v1Coding
35.8% best15 / 61
14.4%31 / 61
60.4% best4 / 51
0.0%51 / 51