Inference
EN

GPT-6 Sol vs MiniMax M3.1 Flash

Change models
GPT-6 SolOpenAI
MiniMax M3.1 FlashMiniMax
IndexAll categories, out of 100.
72.8 ±2.2 best
69.3 ±5.4
GapIndex points below the top.
Top
−3.5 · no clear gap
CoverageIndex weight that has results.
100% best
80%
SpeedOutput tokens per second.
61/s
100/s best
EloLMArena rating and rank.
1397
—
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.
$10.00
$0.60 best
Batch output / 1MBatch price per 1M output tokens.
$5.00
—
1K chats2K tokens in, 500 out, each.
$9.00
$0.60 best
1K documents20K tokens in, 1K out, each.
$50.00
$3.60 best
1K agent tasks200K tokens in, 20K out, each.
$600.00
$42.00 best
ContextMaximum tokens in one request.
1.05M
1.05M
AgenticMulti-step tasks with tools.
68.6
71.8 best
CodingCode writing and repair.
73.9 best
69.4
ReasoningLogic problems and puzzles.
76.1 best
67.9
MultimodalTasks with images and text.
70.1 best
58.1
KnowledgeFacts and expert knowledge.
71.9 best
68.9
MultilingualTasks in many languages.
65.5
—
InstructionTasks with strict prompt rules.
65.3
—
MathMath problems.
75.8
—

Benchmarks

24 shared
Only GPT-6 Sol: 70Only MiniMax M3.1 Flash: 12
GPT-6 Sol
MiniMax M3.1 Flash
Artificial Analysis Humanity's Last ExamKnowledge
47.9%18 / 222
48.0% best17 / 222
Artificial Analysis Long Context ReasoningReasoning
83.7% best10 / 217
79.7%45 / 217
Critical Physics TasksReasoning
30.9% best8 / 214
22.3%25 / 214
Artificial Analysis Omniscience AccuracyKnowledge
54.5% best13 / 164
45.1%37 / 164
GDPval-AA normalizedAgentic
50.4% best27 / 133
47.9%31 / 133
Vibe Code Bench v1.1Coding
87.8% best12 / 123
73.9%45 / 123
Artificial Analysis SciCodeCoding
57.6% best18 / 114
53.4%39 / 114
Artificial Analysis MMMU-ProMultimodal
82.9% best12 / 113
79.1%29 / 113
Legal Research BenchLegal
28.8%57 / 90
34.1% best50 / 90
Tax Agent BenchFinance
15.0%62 / 82
32.1% best23 / 82
LiveBench Data AnalysisReasoning
81.2% best7 / 80
71.4%52 / 80
LiveBench ReasoningReasoning
88.7% best20 / 80
82.9%48 / 80
Terminal-Bench 4.0Agentic
44.4% best13 / 61
25.3%28 / 61
DeepSWEAgentic
68.8%24 / 59
73.7% best13 / 59
Agent Arena steerabilityAgentic
10.1 best11 / 54
−2.636 / 54
Agent Arena task outcomeAgentic
4.5 best17 / 54
−0.734 / 54
LiveBench Coding 2.0Coding
93.1% best9 / 52
83.9%26 / 52
LiveBench DesignAgentic
82.2% best9 / 52
74.0%31 / 52
LiveBench ResearchAgentic
69.3%34 / 52
72.7% best29 / 52
LiveBench Long AgenticAgentic
66.9% best6 / 51
34.7%29 / 51