Model comparison
MiniMax-M3 vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 43.8 on the Noometry Index. MiniMax-M3 costs 5.7× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Last verified . 31 shared benchmarks.
MiniMax
Alibaba (Qwen)Summary
- They share 31 benchmarks with published results for both. MiniMax-M3 scores higher in 1 category and Qwen3.8 Max in 9 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3.8 Max leads 73.2 to 40.0.
- The biggest single-benchmark swing is ProofBench: 18% for MiniMax-M3 and 58% for Qwen3.8 Max.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- MiniMax-M3 has downloadable open weights; the other is API-only.
Side by side
| MiniMax-M3 | Qwen3.8 Max | |
|---|---|---|
| Provider | MiniMax | Alibaba (Qwen) |
| Noometry Index | 43.8 | 56.8 |
| Released | 2026-06-01 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 1M | 1M |
| Max output | 512K | 131K |
| Input $ / M tokens | $0.30 | $2 |
| Output $ / M tokens | $1.20 | $6 |
| Results tracked | 41 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
MiniMax-M3: 41.8 (#118), Qwen3.8 Max: 53.5 (#29)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1482 | 1674 |
| SciCode | 47.1% | 53.2% |
| LMArena Coding | 1469 | 1502 |
| DeepSWE | — | 57.5% |
| FrontierCode | 14.7% | — |
| FrontierSWE | — | 17.8% |
| ALE-Bench | 640.02 | — |
Agentic & Tool Use Qwen3.8 Max leads
MiniMax-M3: 22.6 (#130), Qwen3.8 Max: 45.4 (#14)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | 37.7% | 63.3% |
| OSWorld 2.0 | 4.6% | — |
| τ²-bench Banking | — | 55.1% |
| GBAEval | 0.9% | — |
| GDP.pdf | — | 23.2% |
| Vending-Bench 2 | 2,158 | — |
Reasoning Qwen3.8 Max leads
MiniMax-M3: 30.1 (#87), Qwen3.8 Max: 54.4 (#26)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 65.1% | 88.3% |
| CritPt | 3.7% | 20% |
| Chess Puzzles | 14% | 40% |
| LMArena Hard Prompts | 1447 | 1496 |
| Mystery Game Puzzles | 8% | 38% |
| DTBench | 78.9% | 92% |
| LMCA | 33.7% | 46.2% |
| Epoch Capabilities Index | 146.95 | 156.41 |
| SimpleBench | 45.8% | — |
| Surface Evolver Bench | 55% | — |
| ForecastBench | 61.4 | — |
Math Qwen3.8 Max leads
MiniMax-M3: 40.0 (#95), Qwen3.8 Max: 73.2 (#20)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 71.1% | 100% |
| ProofBench | 18% | 58% |
| LMArena Math | 1429 | 1499 |
| FrontierMath (Tiers 1-3) | — | 74.7% |
| FrontierMath Tier 4 | — | 46.3% |
Knowledge Qwen3.8 Max leads
MiniMax-M3: 58.4 (#35), Qwen3.8 Max: 61.7 (#27)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 90.9% | 92.7% |
| LMArena Expert | 1461 | 1507 |
| SimpleQA Verified | — | 47.3% |
Multimodal MiniMax-M3 leads
MiniMax-M3: 40.2 (#51), Qwen3.8 Max: 37.2 (#75)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | 1253 | 1314 |
| Furniture Assembly | — | 20% |
| LMArena Document | 1435 | — |
Multilingual Qwen3.8 Max leads
MiniMax-M3: 53.0 (#75), Qwen3.8 Max: 56.7 (#18)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1420 | 1472 |
| LMArena Chinese | 1463 | 1538 |
| LMArena French | 1447 | 1503 |
| LMArena German | 1426 | 1483 |
| LMArena Japanese | 1381 | 1467 |
| LMArena Korean | 1372 | 1461 |
| LMArena Russian | 1428 | 1481 |
| LMArena Spanish | 1432 | 1492 |
Instruction Following Qwen3.8 Max leads
MiniMax-M3: 75.5 (#62), Qwen3.8 Max: 77.6 (#17)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1433 | 1479 |
Long Context Qwen3.8 Max leads
MiniMax-M3: 44.2 (#72), Qwen3.8 Max: 45.6 (#31)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1445 | 1489 |
Writing & Preference Qwen3.8 Max leads
MiniMax-M3: 62.1 (#83), Qwen3.8 Max: 67.1 (#30)
| Benchmark | MiniMax-M3 | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1433 | 1483 |
| LMArena Creative Writing | 1404 | 1479 |
| LMArena Multi-Turn | 1442 | 1489 |
| EQ-Bench 4 | 1150 | — |
Frequently asked questions
Is MiniMax-M3 better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 43.8 on the Noometry Index. MiniMax-M3 costs 5.7× less per token, which makes it the better buy when Qwen3.8 Max's lead doesn't matter for your workload.
Which is cheaper, MiniMax-M3 or Qwen3.8 Max?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; Qwen3.8 Max lists at $2 and $6.
Is MiniMax-M3 or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 41.8 in the Noometry coding category.
Which has the bigger context window?
Both accept 1M tokens.
How many benchmarks do MiniMax-M3 and Qwen3.8 Max share?
31 benchmarks have published results for both models. MiniMax-M3 has 41 scored results on Noometry and Qwen3.8 Max has 39.