MiniMax M3 vs Step 5 Preview
Step 5 Preview leads MiniMax M3 on coding, 61.1 to 39.2 on the Vibe Coding Index, and MiniMax M3 costs 2.3x less per output token ($1.20 vs $2.70 per million). Updated October 21, 2026; 55 benchmarks have a result for both.
MiniMax M3 or Step 5 Preview: which to pick
Pick Step 5 Preview for coding: it scores 61.1 on the Vibe Coding Index against 39.2 for MiniMax M3, though on 4 benchmarks against 6.
Pick MiniMax M3 for output tokens 2.3x cheaper ($1.20 vs $2.70 per million) and work measured by LMArena Text: hard prompts in English or LMArena Vision: homework, where it leads.
- Shared results
- 55
- MiniMax M3 only
- 61
- Step 5 Preview only
- 6
- Like-for-like categories
- 8 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
Step 5 Preview
Step 5 Preview scores 61.1 against 39.2.
6 and 4 benchmarks behind the scores
Intelligence
AA Intelligence Index
Step 5 Preview
Step 5 Preview leads, 43.7 against 29.2.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only MiniMax M3 has a AA Agentic Index result, 29.5.
Directional only
Reasoning
GPQA Diamond
No call
Only MiniMax M3 has a GPQA Diamond result, 92.7%.
Directional only
Knowledge
MMLU Pro
No call
Only MiniMax M3 has a MMLU Pro result, 84.2%.
Directional only
Web design
Design Arena: websites
No call
Only MiniMax M3 has a Design Arena: websites result, 1257.
Directional only
Long documents
Listed context window
No call
Both take 1.05M tokens.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
MiniMax M3
MiniMax M3 $0.0009, Step 5 Preview $0.0024 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only MiniMax M3 has the number, $0.56.
Directional only
At a glance
- Coding index
- MiniMax M3 39.2 from 6 benchmarks, Step 5 Preview 61.1 from 4 benchmarks.
- Widest category
- Coding, 15.0 points apart on average over 12 shared benchmarks.
- Widest single gap
- Terminal-Bench 4 (Vals AI): MiniMax M3 1.0% against Step 5 Preview 31.8%, 41.1 points apart on a 0 to 100 scale.
- Output price
- MiniMax M3 $1.20 against Step 5 Preview $2.70 per million output tokens, 2.3x apart.
- Context window
- Both take 1.05M tokens.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | MiniMax M3 | Bars | Step 5 Preview | Ties |
|---|---|---|---|---|
| Coding | 0 | 12 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 7 | ||
| Reasoning | 2 | 5 | ||
| Math | 0 | 3 | ||
| Knowledge | 1 | 3 | ||
| Web design | 0 | 1 | ||
| General | 4 | 16 | ||
| All shared boards | 7 | 48 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index012
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| LiveBench coding | 70.9% | 85.6% | Step 5 Preview |
| LMArena WebDev | 1482 | 1567 | Step 5 Preview |
| AA Coding Index | 58.6 | Not measured | Only MiniMax M3 measured |
| LiveCodeBench (Vals AI) | 82.2% | Not measured | Only MiniMax M3 measured |
| SWE-bench Verified (Vals AI) | 75.0% | Not measured | Only MiniMax M3 measured |
| Terminal-Bench 4 (Vals AI) | 1.0% | 31.8% | Step 5 Preview |
| Vibe Code Bench (Vals AI) | 47.6% | 70.4% | Step 5 Preview |
| Code migration (Vals AI) | 19.9% | 39.0% | Step 5 Preview |
| IOI (Vals AI) | Not measured | 38.1% | Only Step 5 Preview measured |
| Terminal-Bench 2.1 (Vals AI) | 53.6% | Not measured | Only MiniMax M3 measured |
| SWE-bench Multimodal (Vals AI) | 33.3% | 40.8% | Step 5 Preview |
| Vibe Code Bench 1-100 (Vals AI) | 9.2% | Not measured | Only MiniMax M3 measured |
| SkillsBench (Vals AI) | 51.5% | Not measured | Only MiniMax M3 measured |
| CyberBench (Vals AI) | 65.0% | Not measured | Only MiniMax M3 measured |
| SciCode | 47.1% | 58.9% | Step 5 Preview |
| Terminal-Bench Hard (Artificial Analysis run) | 42.4% | Not measured | Only MiniMax M3 measured |
| FrontierCode | 14.7% | Not measured | Only MiniMax M3 measured |
| ALE-Bench | 640.0 | Not measured | Only MiniMax M3 measured |
| LMArena Text: coding | 1470 | 1485 | Step 5 Preview |
| LMArena Text: software and IT | 1466 | 1473 | Step 5 Preview |
| LMArena WebDev: React | 1478 | 1583 | Step 5 Preview |
| LMArena WebDev: HTML | 1455 | 1556 | Step 5 Preview |
| LMArena WebDev: from an image | 1470 | 1550 | Step 5 Preview |
| Neither model has a result on the other 16 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| AA Intelligence Index | 29.2 | 43.7 | Step 5 Preview |
| Epoch Capabilities Index | 146.9 | Not measured | Only MiniMax M3 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index07
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| AA Agentic Index | 29.5 | Not measured | Only MiniMax M3 measured |
| Design Arena: agents building web apps | 1200 | Not measured | Only MiniMax M3 measured |
| Design Arena: agents building full-stack apps | 1197 | Not measured | Only MiniMax M3 measured |
| Design Arena: agents building mobile apps | 1169 | Not measured | Only MiniMax M3 measured |
| Design Arena: agents building Android apps | 1157 | Not measured | Only MiniMax M3 measured |
| Design Arena: agents building games | 1158 | Not measured | Only MiniMax M3 measured |
| Design Arena: HTML slides | 1180 | Not measured | Only MiniMax M3 measured |
| Design Arena: PowerPoint slides | 1205 | Not measured | Only MiniMax M3 measured |
| Design Arena: spreadsheets | 1009 | Not measured | Only MiniMax M3 measured |
| LiveBench research | 68.0% | 72.8% | Step 5 Preview |
| LiveBench browser | 8.0% | 24.0% | Step 5 Preview |
| LiveBench long agentic | 12.4% | 27.5% | Step 5 Preview |
| SREBench (Vals AI) | Not measured | 0.4% | Only Step 5 Preview measured |
| Finance Agent (Vals AI) | 48.3% | 50.7% | Step 5 Preview |
| Tax Agent Bench (Vals AI) | 20.6% | 27.3% | Step 5 Preview |
| Legal Research Bench (Vals AI) | 29.8% | 40.4% | Step 5 Preview |
| Excel Modeling Benchmark (Vals AI) | 47.8% | 57.1% | Step 5 Preview |
| BioMysteryBench (Vals AI) | Not measured | 67.4% | Only Step 5 Preview measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 2.9% | Only Step 5 Preview measured |
| Vals Science Index | Not measured | 29.6% | Only Step 5 Preview measured |
| τ²-Bench Telecom | 88.9% | Not measured | Only MiniMax M3 measured |
| APEX-Agents | 37.7% | Not measured | Only MiniMax M3 measured |
| OSWorld 2.0 | 4.6% | Not measured | Only MiniMax M3 measured |
| Vending-Bench 2 | $2,158 | Not measured | Only MiniMax M3 measured |
| Neither model has a result on the other 17 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index25
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| GPQA Diamond | 92.7% | Not measured | Only MiniMax M3 measured |
| LiveBench reasoning | 74.5% | 83.8% | Step 5 Preview |
| MysteryMechanism (Vals AI) | Not measured | 18.5% | Only Step 5 Preview measured |
| MMMU Pro (Vals AI) | 81.2% | Not measured | Only MiniMax M3 measured |
| GPQA Diamond (Artificial Analysis run) | 92.9% | Not measured | Only MiniMax M3 measured |
| Humanity's Last Exam (Artificial Analysis run) | 39.0% | 46.5% | Step 5 Preview |
| GPQA Diamond (Epoch AI run) | 90.9% | Not measured | Only MiniMax M3 measured |
| SimpleBench | 45.8% | Not measured | Only MiniMax M3 measured |
| Chess puzzles | 13.0% | Not measured | Only MiniMax M3 measured |
| Mystery game puzzles | 8.0% | Not measured | Only MiniMax M3 measured |
| LMArena Text: hard prompts | 1447 | 1461 | Step 5 Preview |
| LMArena Text: expert | 1461 | 1485 | Step 5 Preview |
| LMArena Text: hard prompts in English | 1455 | 1454 | MiniMax M3 |
| LMArena Vision: diagrams | 1266 | 1308 | Step 5 Preview |
| LMArena Vision: homework | 1268 | 1265 | MiniMax M3 |
| Neither model has a result on the other 7 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index03
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| OTIS Mock AIME | 71.1% | Not measured | Only MiniMax M3 measured |
| ProofBench | 18.0% | 42.0% | Step 5 Preview |
| LMArena Text: math | 1429 | 1458 | Step 5 Preview |
| LMArena Text: mathematical work | 1448 | 1460 | Step 5 Preview |
| Neither model has a result on the other 9 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index13
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| MMLU Pro | 84.2% | Not measured | Only MiniMax M3 measured |
| LegalBench (Vals AI) | 85.4% | Not measured | Only MiniMax M3 measured |
| CorpFin (Vals AI) | 68.1% | Not measured | Only MiniMax M3 measured |
| TaxEval (Vals AI) | 72.7% | Not measured | Only MiniMax M3 measured |
| MortgageTax (Vals AI) | 68.4% | Not measured | Only MiniMax M3 measured |
| MedCode (Vals AI) | 46.3% | Not measured | Only MiniMax M3 measured |
| LMArena Text: science | 1452 | 1454 | Step 5 Preview |
| LMArena Text: medicine and healthcare | 1441 | 1439 | MiniMax M3 |
| LMArena Text: legal and government | 1437 | 1437 | Step 5 Preview |
| LMArena Text: business and finance | 1434 | 1435 | Step 5 Preview |
| Neither model has a result on the other 8 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| Design Arena: websites | 1257 | Not measured | Only MiniMax M3 measured |
| Design Arena: UI components | 1245 | Not measured | Only MiniMax M3 measured |
| Design Arena: game dev | 1211 | Not measured | Only MiniMax M3 measured |
| Design Arena: data viz | 1235 | Not measured | Only MiniMax M3 measured |
| Design Arena: 3D | 1209 | Not measured | Only MiniMax M3 measured |
| Design Arena: all code categories | 1247 | Not measured | Only MiniMax M3 measured |
| Design Arena: SVG | 1162 | Not measured | Only MiniMax M3 measured |
| Design Arena: ASCII art | 1161 | Not measured | Only MiniMax M3 measured |
| LiveBench design | 66.2% | 76.9% | Step 5 Preview |
GeneralLike-for-likePublic boards, shown as published, kept out of the index416
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| LiveBench data analysis | 76.2% | 63.5% | MiniMax M3 |
| Public Benefits Bench (Vals AI) | 64.1% | 58.1% | MiniMax M3 |
| SAGE (Vals AI) | 50.6% | Not measured | Only MiniMax M3 measured |
| MedScribe (Vals AI) | 87.3% | Not measured | Only MiniMax M3 measured |
| Vals Multimodal Index | 60.0% | Not measured | Only MiniMax M3 measured |
| IFBench | 82.9% | Not measured | Only MiniMax M3 measured |
| Long context reasoning (AA-LCR) | 83.0% | 88.3% | Step 5 Preview |
| DTBench | 78.9% | Not measured | Only MiniMax M3 measured |
| LMCA | 33.7% | Not measured | Only MiniMax M3 measured |
| Surface Evolver Bench | 55.0% | Not measured | Only MiniMax M3 measured |
| LMArena Text | 1433 | 1440 | Step 5 Preview |
| LMArena Text: instruction following | 1433 | 1438 | Step 5 Preview |
| LMArena Text: creative writing | 1404 | 1409 | Step 5 Preview |
| LMArena Vision | 1251 | 1265 | Step 5 Preview |
| LMArena Text: English | 1443 | 1447 | Step 5 Preview |
| LMArena Text: non-English | 1420 | 1428 | Step 5 Preview |
| LMArena Text: Chinese | 1466 | 1500 | Step 5 Preview |
| LMArena Text: German | 1428 | Not measured | Only MiniMax M3 measured |
| LMArena Text: Japanese | 1380 | 1383 | Step 5 Preview |
| LMArena Text: Korean | 1373 | Not measured | Only MiniMax M3 measured |
| LMArena Text: French | 1444 | 1438 | MiniMax M3 |
| LMArena Text: Spanish | 1433 | 1441 | Step 5 Preview |
| LMArena Text: Russian | 1426 | 1433 | Step 5 Preview |
| LMArena Text: Polish | 1423 | Not measured | Only MiniMax M3 measured |
| LMArena Text: multi-turn | 1443 | 1437 | MiniMax M3 |
| LMArena Text: longer queries | 1446 | 1449 | Step 5 Preview |
| LMArena Text: writing and language | 1419 | 1425 | Step 5 Preview |
| LMArena Text: entertainment and media | 1403 | 1412 | Step 5 Preview |
| LMArena Vision: English | 1263 | 1290 | Step 5 Preview |
| LMArena Vision: Chinese | 1311 | Not measured | Only MiniMax M3 measured |
| LMArena Vision: reading text | 1259 | 1282 | Step 5 Preview |
| LMArena Vision: humor | 1235 | Not measured | Only MiniMax M3 measured |
| LMArena Vision: recognizing things | 1197 | Not measured | Only MiniMax M3 measured |
| LMArena Vision: creative writing | 1234 | Not measured | Only MiniMax M3 measured |
| LMArena Document | 1435 | Not measured | Only MiniMax M3 measured |
| Neither model has a result on the other 11 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | MiniMax M3 | Step 5 Preview | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
MiniMax M3 costs less
Repository review
50K input + 3K output tokens
MiniMax M3 costs less
Agent loop
200K input + 10K output tokens
MiniMax M3 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | MiniMax M3 | StepFun Step 5 Preview |
|---|---|---|
| Maker | MiniMax | StepFun |
| Weights | Open weights | Open weights |
| Released | May 31, 2026 | October 8, 2026 |
| Context window | 1.05M | 1.05M |
| Input $/Mtok | $0.30 | $1.00 |
| Output $/Mtok | $1.20 | $2.70 |
| Coding index | 39.2 | 61.1 |
| Independent measurements behind it | 6 | 4 |
| Cost per solved task | $0.56 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is MiniMax M3 better than Step 5 Preview for coding?
Pick Step 5 Preview for coding: it scores 61.1 on the Vibe Coding Index against 39.2 for MiniMax M3, though on 4 benchmarks against 6. Pick MiniMax M3 for output tokens 2.3x cheaper ($1.20 vs $2.70 per million) and work measured by LMArena Text: hard prompts in English or LMArena Vision: homework, where it leads.
Which is cheaper, MiniMax M3 or Step 5 Preview?
MiniMax M3 costs $0.30 per million input tokens and $1.20 per million output tokens; Step 5 Preview costs $1.00 and $2.70. Cost per solved task is known for MiniMax M3 only.
Which has the bigger context window, MiniMax M3 or Step 5 Preview?
Neither: both take 1.05M tokens.
How many benchmarks is the MiniMax M3 vs Step 5 Preview comparison based on?
55 benchmarks have a result for both models as of October 21, 2026. MiniMax M3's coding index rests on 6 independent benchmarks and Step 5 Preview's on 4. Where only one model has a result, the page says so instead of naming a winner.