Model comparison
MiniMax-M3 vs Qwen2.5 72B Instruct
MiniMax-M3 is the stronger model overall, scoring 43.5 to 31.5 on the Noometry Index.
Last verified . 23 shared benchmarks.
MiniMax
Alibaba (Qwen)Summary
- They share 23 benchmarks with published results for both. MiniMax-M3 scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where MiniMax-M3 leads 58.4 to 26.9.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 71.1% for MiniMax-M3 and 8.1% for Qwen2.5 72B Instruct.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- MiniMax-M3 accepts more context: 1M tokens versus 131K.
Side by side
| MiniMax-M3 | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | MiniMax | Alibaba (Qwen) |
| Noometry Index | 43.5 | 31.5 |
| Released | 2026-06-01 | 2024-09 |
| Weights | Open | Open |
| Context window | 1M | 131K |
| Max output | 512K | 8K |
| Input $ / M tokens | $0.30 | $1.40 |
| Output $ / M tokens | $1.20 | $5.60 |
| Results tracked | 41 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding MiniMax-M3 leads
MiniMax-M3: 41.5 (#111), Qwen2.5 72B Instruct: 32.8 (#256)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Coding | 1470 | 1293 |
| FrontierCode | 14.7% | — |
| LMArena WebDev | 1482 | — |
| SciCode | 47.1% | — |
| WeirdML | — | 16% |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
| ALE-Bench | 640.02 | — |
Agentic & Tool Use Too close to call
MiniMax-M3: 22.5 (#121), Qwen2.5 72B Instruct: 22.0 (#124)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| APEX-Agents | 37.7% | — |
| OSWorld 2.0 | 4.6% | — |
| TheAgentCompany | — | 5.7% |
| BALROG | — | 16.2% |
| GBAEval | 0.9% | — |
| METR Time Horizons | — | 35.8% |
| Vending-Bench 2 | 2,158 | — |
Reasoning MiniMax-M3 leads
MiniMax-M3: 29.3 (#83), Qwen2.5 72B Instruct: 21.4 (#197)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1447 | 1271 |
| DTBench | 78.9% | 62.9% |
| LMCA | 33.7% | 13.4% |
| Epoch Capabilities Index | 147 | 129.01 |
| ForecastBench | 61.4 | 57.5 |
| SimpleBench | 45.8% | — |
| NYT Connections (extended) | 65.1% | — |
| CritPt | 3.7% | — |
| Chess Puzzles | 14% | — |
| Mystery Game Puzzles | 8% | — |
| Surface Evolver Bench | 55% | — |
| BIG-Bench Hard | — | 79.8% |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math MiniMax-M3 leads
MiniMax-M3: 39.1 (#83), Qwen2.5 72B Instruct: 18.1 (#281)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 71.1% | 8.1% |
| LMArena Math | 1432 | 1283 |
| ProofBench | 18% | — |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge MiniMax-M3 leads
MiniMax-M3: 58.4 (#32), Qwen2.5 72B Instruct: 26.9 (#247)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 90.9% | 49.1% |
| LMArena Expert | 1462 | 1245 |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multimodal Not comparable
MiniMax-M3: 39.5 (#45), Qwen2.5 72B Instruct: —
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Vision | 1254 | — |
| LMArena Document | 1435 | — |
Multilingual MiniMax-M3 leads
MiniMax-M3: 53.2 (#70), Qwen2.5 72B Instruct: 41.2 (#208)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1420 | 1252 |
| LMArena Chinese | 1465 | 1272 |
| LMArena French | 1445 | 1279 |
| LMArena German | 1424 | 1234 |
| LMArena Japanese | 1381 | 1180 |
| LMArena Korean | 1372 | 1188 |
| LMArena Russian | 1430 | 1264 |
| LMArena Spanish | 1431 | 1255 |
Instruction Following MiniMax-M3 leads
MiniMax-M3: 75.7 (#56), Qwen2.5 72B Instruct: 65.7 (#216)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1433 | 1254 |
| IFEval | — | 80.6% |
Long Context MiniMax-M3 leads
MiniMax-M3: 44.2 (#68), Qwen2.5 72B Instruct: 38.9 (#183)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1445 | 1282 |
Writing & Preference MiniMax-M3 leads
MiniMax-M3: 62.3 (#79), Qwen2.5 72B Instruct: 46.9 (#210)
| Benchmark | MiniMax-M3 | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1433 | 1269 |
| LMArena Creative Writing | 1405 | 1221 |
| LMArena Multi-Turn | 1441 | 1272 |
| WildBench | — | 80.2% |
| EQ-Bench 4 | 1150 | — |
Frequently asked questions
Is MiniMax-M3 better than Qwen2.5 72B Instruct?
MiniMax-M3 is the stronger model overall, scoring 43.5 to 31.5 on the Noometry Index.
Which is cheaper, MiniMax-M3 or Qwen2.5 72B Instruct?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is MiniMax-M3 or Qwen2.5 72B Instruct better for coding?
MiniMax-M3 scores higher on coding benchmarks: 41.5 versus 32.8 in the Noometry coding category.
Which has the bigger context window?
MiniMax-M3 does, with 1M tokens against 131K.
How many benchmarks do MiniMax-M3 and Qwen2.5 72B Instruct share?
23 benchmarks have published results for both models. MiniMax-M3 has 41 scored results on Noometry and Qwen2.5 72B Instruct has 43.