Compare / MiniMax M3.1 Flash vs GPT-6 Luna

MiniMax M3.1 Flash vs GPT-6 Luna

Which is cheaper, which is more capable, and which does better at each kind of work, from 90 results on 6 sources that measured both models.

MiniMaxMiniMax M3.1 Flash Released 19 Oct 2026 · $0.15 in · $0.60 out
OpenAIGPT-6 Luna Released 22 Sep 2026 · $0.10 in · $0.50 out
MiniMax M3.1 Flash wins 1517 too close to callGPT-6 Luna wins 58

Of the 90 results both models have, the better value on each. Results whose reported ranges overlap are too close to call.

Results as of 19 November 2026 (catalogue release 2026-11-19-3608133e8331). See also MiniMax vs OpenAI, every model from both labs.

MiniMax M3.1 Flash or GPT-6 Luna? The short answer

They score about the same, so pick GPT-6 Luna: it costs 2.2 times less per survey plan and analysis on SurveyBench.

  • Spring Prompt overall: MiniMax M3.1 Flash 40, GPT-6 Luna 38 (out of 100; how the score works).
  • GPT-6 Luna is cheaper to run: $0.009 against $0.020 per survey plan and analysis on SurveyBench (2.2 times less).
  • MiniMax M3.1 Flash scores higher on the Artificial Analysis Intelligence Index: 42.4 against 38.1.
  • MiniMax M3.1 Flash is newer: released 19 Oct 2026, against 22 Sep 2026 for GPT-6 Luna.
  • On our own benchmarks, on the results both models have: GPT-6 Luna does better for product listings, user surveys and marketing planning.
  • On third-party benchmarks, on the results both models have: MiniMax M3.1 Flash does better for agents and tool use and reasoning and knowledge.

Head to head

40
Spring Prompt overall
38
42.4
Intelligence Index
38.1
$0.26
Price per million tokens, blendedlower is better
$0.20
159
Output speed, tokens a second
127
1,048,576
Context window, tokens
1,050,000

Every fact side by side

MiniMax M3.1 FlashGPT-6 Luna
Price per million tokens$0.15 in · $0.60 out$0.10 in · $0.50 out
Blended (3 in : 1 out)$0.26$0.20
Cheapest host, blended$0.20 (GMICloud, FP8)$0.20 (Amazon Bedrock)
Cost per survey plan and analysis (SurveyBench)$0.020$0.009
Cached input, per million$0.015 cached input$0.010 cached input
Intelligence Index42.438.1
Spring Prompt overall4038
Output speed, tokens a second159127
Context window1,048,576 tokens1,050,000 tokens
Released19 Oct 2026 (newer)22 Sep 2026
WeightsOpenClosed
Developer based inChinathe US

Prices are the developer's list price where we have read it, otherwise the typical host on OpenRouter; see every model's price. Intelligence and speed from Artificial Analysis.

Which is better for what

Each model's rank among the models on sale today, for each kind of work, from our own benchmarks and licensed sources. Further right is better; the better rank is in green.

Product listingsTurning a sparse product feed and photos into listings that can go live 14 of 35 10 of 35
Decks from an analysisTurning a finished analysis into a deck you could present as it is 10 of 35 12 of 35
User surveysPlanning a user survey and reading its results without being misled 26 of 26 14 of 26
Marketing planningPlanning a year of ad spend without overspending 24 of 35 7 of 35
Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls 33 of 159 53 of 159
Professional workReal tasks from banking, consulting and law, business documents and freelance projects not measured 25 of 51
Reasoning and knowledgeHard questions across science, maths and general knowledge 19 of 187 48 of 187
WritingWhat people prefer in blind comparisons, and judged writing quality not measured 61 of 169
Sticking to the factsSummarising without inventing things, and factual answers not measured 89 of 139
Speed under pressureGood decisions against a real clock (fast chess) 34 of 38 28 of 38

Their slides, side by side

The same brief for both: Manchester store investment review for Hearthside Coffee, an invented company. First slide of each deck from DeckBench.

First slide of MiniMax M3.1 Flash's deck for Hearthside Coffee
MiniMax M3.1 Flash · deck rating 1,266 · needs work
First slide of GPT-6 Luna's deck for Hearthside Coffee
GPT-6 Luna · deck rating 1,252 · needs work

Every slide, side by side →

From our research

Every shared result

Each model's best result on every measure both have, grouped by benchmark, with the setting that produced it. Under each measure, every model that benchmark has measured is a grey tick, best to the right, so you can see where the two sit in the field. The headline measures of each benchmark come first; the rest are under "Show all". Green marks the better value; where a source reports ranges that overlap, neither is marked.

MeasureMiniMax M3.1 FlashGPT-6 Luna
BulletBench measured by Spring Prompt · 3 results
Cost per game · Bullet 60sUS dollars, lower is better 20 · 21 of 62 measured $0.0016low reasoning $0.0016no reasoning
Ladder Elo · Bullet 60sladder Elo, higher is better 58 · 52 of 62 measured 115low reasoning 249no reasoning
Games lost on time · Bullet 60s% of games, lower is better 52 · 39 of 62 measured 75.0%low reasoning 50.0%no reasoning
CatalogBench measured by Spring Prompt · 3 results
Cost per productUS dollars, lower is better 1 · 2 of 36 measured $0 $0.0005
Reliably publish-ready% of products, higher is better 17 · 13 of 35 measured 50.0% 55.4%
Reliably publish-ready · sales brief% of products, higher is better 10 · 7 of 35 measured 37.5% 51.8%
DeckBench measured by Spring Prompt · 3 results
Cost per deckUS dollars, lower is better 1 · 3 of 36 measured $0 $0.0087
Deck ratingrating, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 10 · 12 of 36 measured 1,266 1,252
Accurate decks% of tasks, higher is better 6 · 17 of 36 measured 83.3% 50.0%
ROASBench measured by Spring Prompt · 3 results
Overall scorescore out of 100, higher is better 24 · 7 of 36 measured 40.3 53.4
Audience scorescore out of 100, higher is better 22 · 7 of 36 measured 60.8 68.3
Business scorescore out of 100, higher is better 22 · 7 of 36 measured 38.9 54.8
SurveyBench measured by Spring Prompt · 3 results
Analysis ratingrating, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 27 · 14 of 27 measured 428 1,072
Cost per taskUS dollars, lower is better 5 · 1 of 27 measured $0.0198 $0.0088
Analysis scorepoints of 100, higher is better 27 · 7 of 27 measured 13.6 90.8
Artificial Analysis reported by Artificial Analysis · 3 results
Artificial Analysis Intelligence Indexindex score, higher is better 21 · 39 of 249 measured 42.4 38.1max reasoning
Humanity's Last Exam% of questions, higher is better 15 · 53 of 250 measured 48.0% 38.5%max reasoning
Terminal-Bench 4.0% of tasks, higher is better 23 · 37 of 120 measured 25.8% 12.6%max reasoning
Show all 90 results (72 more)
MeasureMiniMax M3.1 FlashGPT-6 Luna
BulletBench measured by Spring Prompt · 2 results
Invalid moves · Bullet 60s% of moves, lower is better 1 · 1 of 62 measured 0.0%low reasoning 0.0%no reasoning
Median move time · Bullet 60smilliseconds, lower is better 56 · 39 of 62 measured 3.0 slow reasoning 1.5 sno reasoning
CatalogBench measured by Spring Prompt · 28 results
Channel compliance% of products, higher is better 1 · 1 of 36 measured 100.0% 100.0%
Channel rules broken% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Channel rules broken · sales brief% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Claims to check% of products, lower is better 1 · 1 of 34 measured 0.0% 0.0%
Claims to check · sales brief% of products, lower is better 1 · 1 of 34 measured 0.0% 0.0%
Conflicts caught% of conflicts, higher is better 1 · 1 of 36 measured 100.0% 100.0%
Content quality% of checks, higher is better 27 · 20 of 36 measured 96.2% 97.0%
Content quality · sales brief% of checks, higher is better 24 · 18 of 36 measured 96.3% 97.4%
Decision accuracy% of decisions, higher is better 24 · 16 of 36 measured 96.1% 97.0%
Failed outputs% of products, lower is better 31 · 1 of 36 measured 1.2% 0.0%
Failed outputs · sales brief% of products, lower is better 31 · 1 of 36 measured 1.2% 0.0%
Field accuracy% of missing fields, higher is better 27 · 18 of 36 measured 92.7% 93.7%
Invented values% of filled values, lower is better 34 · 18 of 36 measured 3.8% 1.7%
Missing UK information% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Missing UK information · sales brief% of products, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Not findable% of products, lower is better 1 · 32 of 36 measured 0.0% 10.1%
Not findable · sales brief% of products, lower is better 1 · 32 of 36 measured 0.0% 10.1%
Publish-ready listings% of products, higher is better 20 · 17 of 36 measured 63.1% 67.9%
Publish-ready listings · sales brief% of products, higher is better 12 · 8 of 36 measured 57.1% 67.9%
Time per productseconds, lower is better 27 · 8 of 36 measured 25.2 8.20
Unsupported claimsclaims per product, lower is better 15 · 13 of 36 measured 0.39 0.31
Unsupported claims% of products, lower is better 12 · 13 of 36 measured 7.7% 8.9%
Unsupported claims · sales briefclaims per product, lower is better 11 · 8 of 36 measured 0.60 0.35
Unsupported claims · sales brief% of products, lower is better 9 · 8 of 36 measured 15.5% 8.3%
Wrong attributes% of products, lower is better 32 · 24 of 36 measured 22.6% 18.5%
Wrong attributes · sales brief% of products, lower is better 30 · 23 of 36 measured 21.4% 19.1%
Wrong category or variant% of products, lower is better 23 · 7 of 36 measured 2.4% 1.8%
Wrong category or variant · sales brief% of products, lower is better 32 · 1 of 36 measured 4.8% 0.0%
DeckBench measured by Spring Prompt · 14 results
Caveat dropped% of tasks, lower is better 16 · 1 of 36 measured 16.7% 0.0%
Clean layout% of tasks, higher is better 31 · 10 of 36 measured 0.0% 66.7%
Design quality% of the maximum, higher is better 19 · 10 of 36 measured 66.9% 78.1%
Draft figure quoted% of tasks, lower is better 1 · 21 of 36 measured 0.0% 33.3%
Findings missing% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other deck was; the win rate counts wins against whichever decks a model happened to meet, so the two can disagree. 14 · 9 of 36 measured 63.9% 70.9%
Layout defects% of tasks, lower is better 31 · 10 of 36 measured 100.0% 33.3%
Misleading metric used% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Presentable decks% of tasks, higher is better 17 · 17 of 36 measured 0.0% 0.0%
Recommendation late or wrong% of tasks, lower is better 1 · 17 of 36 measured 0.0% 16.7%
Slides needing work% of tasks, lower is better 19 · 5 of 36 measured 50.0% 16.7%
Time per deckseconds, lower is better 26 · 17 of 36 measured 217 101
Unsupported claims% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
Unsupported numbers% of tasks, lower is better 1 · 1 of 36 measured 0.0% 0.0%
ROASBench measured by Spring Prompt · 6 results
Consistency scorescore out of 100, higher is better 29 · 10 of 36 measured 17.3 35.8
Contribution profitsimulated US dollars, higher is better 13 · 9 of 36 measured $693,379 $738,879
Cost of a runUS dollars, lower is better 1 · 2 of 36 measured $0 $0.0142
Months over budgetmonths of 12, lower is better 1 · 1 of 36 measured 0 0
Planning scorescore out of 100, higher is better 28 · 20 of 36 measured 54.4 54.8
Return on ad spendprofit per $1 spent, higher is better 19 · 10 of 36 measured $0.52 $0.66
SurveyBench measured by Spring Prompt · 18 results
Head-to-head win rate% of comparisons, higher is betterThe rating weighs each win by how strong the other analysis was; the win rate counts wins against whichever analyses a model happened to meet, so the two can disagree. 27 · 17 of 27 measured 20.3% 47.1%
Leading or double-barrelled questions% of tasks, lower is better 15 · 5 of 27 measured 44.4% 11.1%
Mistaken next steps% of tasks, lower is better 24 · 5 of 27 measured 77.8% 11.1%
Numbers right% of answers, higher is better 1 · 21 of 27 measured 100.0% 98.0%
Objectives unanswerable% of tasks, lower is better 27 · 1 of 27 measured 44.4% 0.0%
Output tokens per tasktokens, lower is better 16 · 4 of 17 measured 63,237 14,333
Precision ignored% of tasks, lower is better 21 · 1 of 27 measured 44.4% 0.0%
Research scorepoints of 100, higher is better 27 · 7 of 27 measured 28.3 91.7
Sound research% of tasks, higher is better 17 · 6 of 27 measured 0.0% 33.3%
Stakeholder's view presumed% of tasks, lower is better 25 · 12 of 27 measured 33.3% 11.1%
Survey plan scorepoints of 100, higher is better 26 · 9 of 27 measured 62.5 94.0
Survey structure% of tasks, lower is better 25 · 12 of 27 measured 44.4% 22.2%
Time per taskseconds, lower is better 21 · 12 of 27 measured 352 138
Traps handled% of traps, higher is better 22 · 11 of 27 measured 75.0% 94.2%
Traps missed% of tasks, lower is better 19 · 8 of 27 measured 88.9% 33.3%
Unsupported findings% of tasks, lower is better 22 · 7 of 27 measured 88.9% 22.2%
Wrong numbers% of tasks, lower is better 24 · 14 of 27 measured 88.9% 11.1%
Wrong recommendation% of tasks, lower is better 25 · 5 of 27 measured 88.9% 11.1%
Artificial Analysis reported by Artificial Analysis · 4 results
Long-context reasoning (AA-LCR)% of questions, higher is better 47 · 13 of 241 measured 79.7% 83.3%max reasoning
Output speedtokens per second, higher is better 21 · 37 of 78 measured 159 127max reasoning
SciCode% of problems, higher is better 38 · 31 of 121 measured 53.4% 54.6%max reasoning
Time to first answer tokenseconds, lower is better 51 · 20 of 78 measured 13.3 0.74no reasoning

Quick answers

Is MiniMax M3.1 Flash better than GPT-6 Luna?

They score about the same, so pick GPT-6 Luna: it costs 2.2 times less per survey plan and analysis on SurveyBench.

Which is cheaper, MiniMax M3.1 Flash or GPT-6 Luna?

GPT-6 Luna is cheaper to run: $0.009 against $0.020 per survey plan and analysis on SurveyBench (2.2 times less).

Which is faster, MiniMax M3.1 Flash or GPT-6 Luna?

MiniMax M3.1 Flash writes about 159 tokens a second on its usual API, against 127 for GPT-6 Luna (Artificial Analysis).

Which is better for product listings, MiniMax M3.1 Flash or GPT-6 Luna?

GPT-6 Luna ranks 10 of 35 models on sale for turning a sparse product feed and photos into listings that can go live, against 14 for MiniMax M3.1 Flash.

Which is better for decks from an analysis, MiniMax M3.1 Flash or GPT-6 Luna?

MiniMax M3.1 Flash ranks 10 of 35 models on sale for turning a finished analysis into a deck you could present as it is, against 12 for GPT-6 Luna.

Which is better for user surveys, MiniMax M3.1 Flash or GPT-6 Luna?

GPT-6 Luna ranks 14 of 26 models on sale for planning a user survey and reading its results without being misled, against 26 for MiniMax M3.1 Flash.

Which is better for marketing planning, MiniMax M3.1 Flash or GPT-6 Luna?

GPT-6 Luna ranks 7 of 35 models on sale for planning a year of ad spend without overspending, against 24 for MiniMax M3.1 Flash.

Which has the bigger context window, MiniMax M3.1 Flash or GPT-6 Luna?

GPT-6 Luna: 1,050,000 tokens, against 1,048,576 for MiniMax M3.1 Flash.

Run MiniMax M3.1 Flash and GPT-6 Luna on your own prompt.

Benchmarks aren't your task. Test both side by side on your prompt, with every answer checked against your rules and the cost of each one.