Compare / MiniMax M3 vs GPT-5.6 Luna
MiniMax M3 vs GPT-5.6 Luna
Which is cheaper, which is more capable, and which does better at each kind of work, from 22 results on 3 sources that measured both models.
MiniMax M3 or GPT-5.6 Luna? The short answer
- GPT-5.6 Luna is cheaper: $0.45 against $0.52 per million tokens, 14% less.
- GPT-5.6 Luna scores higher on the Artificial Analysis Intelligence Index: 37.3 against 29.2.
- On our business benchmarks, MiniMax M3 ranks higher for reasoning and knowledge.
- On our business benchmarks, GPT-5.6 Luna ranks higher for agents and tool use, professional work, writing and sticking to the facts.
At a glance
| MiniMax M3 | GPT-5.6 Luna | |
|---|---|---|
| Price per million tokens | $0.30 in · $1.20 out | $0.20 in · $1.20 out |
| Blended (3 in : 1 out) | $0.52 | $0.45 |
| Cheapest host, blended | $0.41 (CoreWeave, FP4) | $0.23 (OpenAI) |
| Intelligence Index | 29.2 | 37.3 |
| Spring Prompt overall | – | – |
| Output speed, tokens a second | 81 | – |
| Context window | 1,048,576 tokens | 1,050,000 tokens |
| Weights | Open | Closed |
| Developer based in | China | the 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 models available today, for each kind of work, from our own benchmarks and licensed sources. The better rank is in green.
| Use case | MiniMax M3 | GPT-5.6 Luna |
|---|---|---|
| Agents and tool useMulti-step tasks with tools: support desks, coding agents, function calls | 71 of 161 | 47 of 161 |
| Professional workReal tasks from banking, consulting and law, business documents and freelance projects | 39 of 51 | 20 of 51 |
| Reasoning and knowledgeHard questions across science, maths and general knowledge | 36 of 195 | 38 of 195 |
| WritingWhat people prefer in blind comparisons, and judged writing quality | 61 of 187 | 57 of 187 |
| Sticking to the factsSummarising without inventing things, and factual answers | 94 of 153 | 59 of 153 |
Every shared result
| Benchmark | Metric | MiniMax M3 | GPT-5.6 Luna |
|---|---|---|---|
| Arena (formerly LMArena)reported | Confirmed task successIPS effect estimate | -0.14 | -0.05 |
| Arena (formerly LMArena)reported | Praise over complaintIPS effect estimate | -0.13 | -0.04 |
| Arena (formerly LMArena)reported | SteerabilityIPS effect estimate | -0.07 | 0.02 |
| Arena (formerly LMArena)reported | Tool groundingIPS effect estimate | -0.00 | 0.00 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,433 | 1,462 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,434 | 1,461 |
| Arena (formerly LMArena)reported | Business, management and financeArena rating | 1,450 | 1,463 |
| Arena (formerly LMArena)reported | Creative writingArena rating | 1,406 | 1,412 |
| Arena (formerly LMArena)reported | Expert promptsArena rating | 1,471 | 1,491 |
| Arena (formerly LMArena)reported | Instruction followingArena rating | 1,435 | 1,446 |
| Arena (formerly LMArena)reported | OverallArena rating | 1,440 | 1,454 |
| Arena (formerly LMArena)reported | Writing, literature and languageArena rating | 1,422 | 1,435 |
| Artificial Analysisreported | Artificial Analysis Coding Indexindex score | 58.6 | 71.4 |
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 29.2 | 37.3 |
| Artificial Analysisreported | GPQA Diamond% of questions | 92.9% | 91.1% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 39.0% | 39.5% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 83.0% | 83.7% |
| Artificial Analysisreported | SciCode% of problems | 47.1% | 53.6% |
| Artificial Analysisreported | Terminal-Bench 2.1% of tasks | 65.2% | 80.9% |
| Artificial Analysisreported | Terminal-Bench 4.0% of tasks | 2.0% | 11.6% |
| Artificial Analysisreported | Τ-bench banking% of tasks | 15.3% | 31.1% |
| Vending-Bench 2reported | Money after a yearUS dollars | $2,157.77 | $4,094.71 |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.
Quick answers
Is MiniMax M3 better than GPT-5.6 Luna?
It depends on the task. Among models available today, on our benchmarks and licensed sources, MiniMax M3 ranks higher for reasoning and knowledge; GPT-5.6 Luna ranks higher for agents and tool use, professional work, writing and sticking to the facts.
Which is cheaper, MiniMax M3 or GPT-5.6 Luna?
GPT-5.6 Luna costs $0.45 per million tokens (three input to one output), against $0.52 for MiniMax M3.
Run MiniMax M3 and GPT-5.6 Luna on your own prompt
Benchmarks aren't your data. Try both side by side in the playground, with the cost of every answer.