Inference
EN

DeepSeek V4.1 Flash 1017 vs MiniMax M3.1 Flash

Change models
DeepSeek V4.1 Flash 1017DeepSeekFastest
MiniMax M3.1 FlashMiniMax
IndexAll categories, out of 100.
68.5 ±2.6
69.3 ±5.4 best
GapIndex points below the top.
−0.8 · no clear gap
Top
CoverageIndex weight that has results.
95% best
80%
SpeedOutput tokens per second.
261/s best
100/s
EloLMArena rating and rank.
1463
—
ReleasedMonth of release.
2026-10
2026-10
Input / 1MUS dollars per 1M input tokens.
$0.022 best
$0.15
Output / 1MUS dollars per 1M output tokens.
$0.40 best
$0.60
Batch output / 1MBatch price per 1M output tokens.
$0.20
—
1K chats2K tokens in, 500 out, each.
$0.244 best
$0.60
1K documents20K tokens in, 1K out, each.
$0.84 best
$3.60
1K agent tasks200K tokens in, 20K out, each.
$12.40 best
$42.00
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
72.2 best
71.8
CodingCode writing and repair.
69.8 best
69.4
ReasoningLogic problems and puzzles.
65.0
67.9 best
MultimodalTasks with images and text.
62.2 best
58.1
KnowledgeFacts and expert knowledge.
64.2
68.9 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
68.3
—
MathMath problems.
65.4
—

Benchmarks

27 shared
Only DeepSeek V4.1 Flash 1017: 67Only MiniMax M3.1 Flash: 9
DeepSeek V4.1 Flash 1017
MiniMax M3.1 Flash
Artificial Analysis Humanity's Last ExamKnowledge
39.2%51 / 222
48.0% best17 / 222
Artificial Analysis Long Context ReasoningReasoning
84.0% best8 / 217
79.7%45 / 217
Critical Physics TasksReasoning
14.3%52 / 214
22.3% best25 / 214
Artificial Analysis Omniscience AccuracyKnowledge
46.4% best33 / 164
45.1%37 / 164
GDPval-AA normalizedAgentic
55.0% best21 / 133
47.9%31 / 133
Vibe Code Bench v1.1Coding
84.7% best18 / 123
73.9%45 / 123
Artificial Analysis SciCodeCoding
51.9%44 / 114
53.4% best39 / 114
Artificial Analysis MMMU-ProMultimodal
77.0%43 / 113
79.1% best29 / 113
Legal Research BenchLegal
41.3% best30 / 90
34.1%50 / 90
Tax Agent BenchFinance
26.6%41 / 82
32.1% best23 / 82
LiveBench Data AnalysisReasoning
79.3% best15 / 80
71.4%52 / 80
LiveBench ReasoningReasoning
86.7% best31 / 80
82.9%48 / 80
Terminal-Bench 4.0Agentic
19.7%34 / 61
25.3% best28 / 61
DeepSWEAgentic
77.0% best4 / 59
73.7%13 / 59
Agent Arena steerabilityAgentic
4.0 best18 / 54
−2.636 / 54
Agent Arena task outcomeAgentic
2.3 best25 / 54
−0.734 / 54
LiveBench Coding 2.0Coding
90.2% best13 / 52
83.9%26 / 52
LiveBench DesignAgentic
79.2% best15 / 52
74.0%31 / 52
LiveBench ResearchAgentic
86.8% best4 / 52
72.7%29 / 52
LiveBench Long AgenticAgentic
40.1% best22 / 51
34.7%29 / 51