Inference
EN

Grok 4.7 vs MiniMax M3.1 Flash

Change models
Grok 4.7xAI
MiniMax M3.1 FlashMiniMax
IndexAll categories, out of 100.
69.1 ±3.7
69.3 ±5.4 best
GapIndex points below the top.
−0.2 · no clear gap
Top
CoverageIndex weight that has results.
85% best
80%
SpeedOutput tokens per second.
51/s
100/s best
EloLMArena rating and rank.
1398
—
ReleasedMonth of release.
2026-09
2026-10
Input / 1MUS dollars per 1M input tokens.
$2.00
$0.15 best
Output / 1MUS dollars per 1M output tokens.
$6.00
$0.60 best
Batch output / 1MBatch price per 1M output tokens.
$4.80
—
1K chats2K tokens in, 500 out, each.
$7.00
$0.60 best
1K documents20K tokens in, 1K out, each.
$46.00
$3.60 best
1K agent tasks200K tokens in, 20K out, each.
$520.00
$42.00 best
ContextMaximum tokens in one request.
500K
1.05M best
AgenticMulti-step tasks with tools.
71.9 best
71.8
CodingCode writing and repair.
69.9 best
69.4
ReasoningLogic problems and puzzles.
64.5
67.9 best
MultimodalTasks with images and text.
—
58.1
KnowledgeFacts and expert knowledge.
66.5
68.9 best
MultilingualTasks in many languages.
—
—
InstructionTasks with strict prompt rules.
78.9
—
MathMath problems.
61.8
—

Benchmarks

23 shared
Only Grok 4.7: 61Only MiniMax M3.1 Flash: 13
Grok 4.7
MiniMax M3.1 Flash
Artificial Analysis Humanity's Last ExamKnowledge
43.1%32 / 222
48.0% best17 / 222
Artificial Analysis Long Context ReasoningReasoning
76.7%84 / 217
79.7% best45 / 217
Critical Physics TasksReasoning
17.7%37 / 214
22.3% best25 / 214
Artificial Analysis Omniscience AccuracyKnowledge
42.7%42 / 164
45.1% best37 / 164
GDPval-AA normalizedAgentic
60.7% best8 / 133
47.9%31 / 133
Vibe Code Bench v1.1Coding
86.2% best13 / 123
73.9%45 / 123
Artificial Analysis SciCodeCoding
57.4% best19 / 114
53.4%39 / 114
Legal Research BenchLegal
47.1% best16 / 90
34.1%50 / 90
Tax Agent BenchFinance
35.8% best16 / 82
32.1%23 / 82
LiveBench Data AnalysisReasoning
76.9% best30 / 80
71.4%52 / 80
LiveBench ReasoningReasoning
82.7%51 / 80
82.9% best48 / 80
Terminal-Bench 4.0Agentic
28.8% best25 / 61
25.3%28 / 61
DeepSWEAgentic
71.0%21 / 59
73.7% best13 / 59
Agent Arena steerabilityAgentic
−2.3 best35 / 54
−2.636 / 54
Agent Arena task outcomeAgentic
5.1 best14 / 54
−0.734 / 54
LiveBench Coding 2.0Coding
84.5% best24 / 52
83.9%26 / 52
LiveBench DesignAgentic
78.0% best20 / 52
74.0%31 / 52
LiveBench ResearchAgentic
86.4% best5 / 52
72.7%29 / 52
LiveBench Long AgenticAgentic
50.1% best11 / 51
34.7%29 / 51
LiveBench Browser UseAgentic
60.0%28 / 49
92.0% best4 / 49