Inference
EN

DeepSeek V4.1 Flash 1017 vs Step 5 Preview

Change models
DeepSeek V4.1 Flash 1017DeepSeekFastest
Step 5 PreviewStepFun
IndexAll categories, out of 100.
68.5 ±2.6 best
68.4 ±3.7
GapIndex points below the top.
Top
−0.1 · no clear gap
CoverageIndex weight that has results.
95% best
90%
SpeedOutput tokens per second.
261/s best
105/s
EloLMArena rating and rank.
1463 best
1443
ReleasedMonth of release.
2026-10
2026-09
Input / 1MUS dollars per 1M input tokens.
$0.022 best
$1.00
Output / 1MUS dollars per 1M output tokens.
$0.40 best
$2.70
Batch output / 1MBatch price per 1M output tokens.
$0.20
—
1K chats2K tokens in, 500 out, each.
$0.244 best
$3.35
1K documents20K tokens in, 1K out, each.
$0.84 best
$22.70
1K agent tasks200K tokens in, 20K out, each.
$12.40 best
$254.00
ContextMaximum tokens in one request.
1.05M best
1M
AgenticMulti-step tasks with tools.
72.2 best
68.8
CodingCode writing and repair.
69.8
71.4 best
ReasoningLogic problems and puzzles.
65.0
67.4 best
MultimodalTasks with images and text.
62.2
62.2
KnowledgeFacts and expert knowledge.
64.2 best
63.5
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
68.3
—
MathMath problems.
65.4 best
61.7

Benchmarks

44 shared
Only DeepSeek V4.1 Flash 1017: 50Only Step 5 Preview: 18
DeepSeek V4.1 Flash 1017
Step 5 Preview
Artificial Analysis Humanity's Last ExamKnowledge
39.2%51 / 222
46.5% best23 / 222
Artificial Analysis Long Context ReasoningReasoning
84.0%8 / 217
88.3% best2 / 217
Critical Physics TasksReasoning
14.3%52 / 214
20.9% best26 / 214
Artificial Analysis Omniscience AccuracyKnowledge
46.4% best33 / 164
36.4%64 / 164
GDPval-AA normalizedAgentic
55.0% best21 / 133
54.3%24 / 133
Vibe Code Bench v1.1Coding
84.7% best18 / 123
70.4%48 / 123
Artificial Analysis SciCodeCoding
51.9%44 / 114
58.9% best13 / 114
Artificial Analysis MMMU-ProMultimodal
77.0% best43 / 113
76.4%44 / 113
Legal Research BenchLegal
41.3% best30 / 90
40.4%35 / 90
Terminal-Bench 2.1 · BenchLMAgentic
90.6% best3 / 86
85.0%20 / 86
GPQA Diamond · BenchLMKnowledge
90.9%24 / 84
93.5% best10 / 84
Tax Agent BenchFinance
26.6%41 / 82
27.3% best39 / 82
LiveBench Data AnalysisReasoning
79.3% best15 / 80
63.5%69 / 80
LiveBench ReasoningReasoning
86.7% best31 / 80
83.8%44 / 80
Finance Agent (v2)Finance
53.5% best33 / 79
50.7%42 / 79
Code MigrationCoding
45.6% best17 / 77
39.0%31 / 77
Humanity's Last ExamKnowledge
42.6%31 / 77
46.5% best26 / 77
EMBFinance
57.2% best40 / 75
57.1%41 / 75
Terminal-Bench 4.0Agentic
19.7%34 / 61
31.8% best23 / 61
DeepSWEAgentic
77.0% best4 / 59
67.7%27 / 59