Model comparison
MiniMax-M2.5 vs Qwen3.8 Max
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 38.3 on the Noometry Index. MiniMax-M2.5 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 . 21 shared benchmarks.
MiniMax
Alibaba (Qwen)Summary
- They share 21 benchmarks with published results for both. MiniMax-M2.5 scores higher in 0 categories and Qwen3.8 Max in 9 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3.8 Max leads 73.2 to 26.9.
- The biggest single-benchmark swing is NYT Connections (extended): 16.8% for MiniMax-M2.5 and 88.3% for Qwen3.8 Max.
- MiniMax-M2.5 is cheaper at $0.30 / $1.20 per million input/output tokens, against $2 / $6 for Qwen3.8 Max.
- Qwen3.8 Max accepts more context: 1M tokens versus 205K.
- MiniMax-M2.5 has downloadable open weights; the other is API-only.
Side by side
| MiniMax-M2.5 | Qwen3.8 Max | |
|---|---|---|
| Provider | MiniMax | Alibaba (Qwen) |
| Noometry Index | 38.3 | 56.8 |
| Released | 2026-02-12 | 2026-08-02 |
| Weights | Open | Proprietary |
| Context window | 205K | 1M |
| Max output | 131K | 131K |
| Input $ / M tokens | $0.30 | $2 |
| Output $ / M tokens | $1.20 | $6 |
| Results tracked | 33 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.8 Max leads
MiniMax-M2.5: 48.1 (#58), Qwen3.8 Max: 53.5 (#29)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena WebDev | 1387 | 1674 |
| LMArena Coding | 1381 | 1502 |
| DeepSWE | — | 57.5% |
| SWE-bench Verified (bash only) | 75.8% | — |
| SWE-bench Multilingual | 68.3% | — |
| FrontierSWE | — | 17.8% |
| SciCode | — | 53.2% |
| ALE-Bench | 618.17 | — |
Agentic & Tool Use Qwen3.8 Max leads
MiniMax-M2.5: 30.4 (#77), Qwen3.8 Max: 45.4 (#14)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| Terminal-Bench | 42.7% | — |
| APEX-Agents | — | 63.3% |
| τ²-bench Banking | — | 55.1% |
| GDP.pdf | — | 23.2% |
| Vending-Bench 2 | -23.16 | — |
Reasoning Qwen3.8 Max leads
MiniMax-M2.5: 17.5 (#292), Qwen3.8 Max: 54.4 (#26)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| NYT Connections (extended) | 16.8% | 88.3% |
| LMArena Hard Prompts | 1372 | 1496 |
| Epoch Capabilities Index | 146.68 | 156.41 |
| ARC-AGI-2 | 4.9% | — |
| Kagi LLM Benchmark | 55.2% | — |
| ARC-AGI-1 | 63.7% | — |
| CritPt | — | 20% |
| Chess Puzzles | — | 40% |
| Mystery Game Puzzles | — | 38% |
| DTBench | — | 92% |
| LMCA | — | 46.2% |
Math Qwen3.8 Max leads
MiniMax-M2.5: 26.9 (#253), Qwen3.8 Max: 73.2 (#20)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| ProofBench | 4% | 58% |
| LMArena Math | 1378 | 1499 |
| FrontierMath (Tiers 1-3) | — | 74.7% |
| FrontierMath Tier 4 | — | 46.3% |
| OTIS Mock AIME 2024-2025 | — | 100% |
Knowledge Qwen3.8 Max leads
MiniMax-M2.5: 39.2 (#135), Qwen3.8 Max: 61.7 (#27)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Expert | 1379 | 1507 |
| GPQA Diamond | — | 92.7% |
| SimpleQA Verified | — | 47.3% |
| Vectara Hallucination Rate | 9.1% | — |
Multimodal Not comparable
MiniMax-M2.5: —, Qwen3.8 Max: 37.2 (#75)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | — | 1314 |
| Furniture Assembly | — | 20% |
Multilingual Qwen3.8 Max leads
MiniMax-M2.5: 47.1 (#152), Qwen3.8 Max: 56.7 (#18)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1338 | 1472 |
| LMArena Chinese | 1393 | 1538 |
| LMArena French | 1362 | 1503 |
| LMArena German | 1362 | 1483 |
| LMArena Japanese | 1171 | 1467 |
| LMArena Korean | 1232 | 1461 |
| LMArena Russian | 1358 | 1481 |
| LMArena Spanish | 1354 | 1492 |
Instruction Following Qwen3.8 Max leads
MiniMax-M2.5: 71.5 (#148), Qwen3.8 Max: 77.6 (#17)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1353 | 1479 |
Long Context Qwen3.8 Max leads
MiniMax-M2.5: 37.5 (#216), Qwen3.8 Max: 45.6 (#31)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1366 | 1489 |
| CL-bench | 11.4% | — |
| CL-bench Life | 6.3% | — |
Writing & Preference Qwen3.8 Max leads
MiniMax-M2.5: 53.9 (#153), Qwen3.8 Max: 67.1 (#30)
| Benchmark | MiniMax-M2.5 | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1359 | 1483 |
| LMArena Creative Writing | 1331 | 1479 |
| LMArena Multi-Turn | 1364 | 1489 |
| EQ-Bench Creative Writing | 1361 | — |
Frequently asked questions
Is MiniMax-M2.5 better than Qwen3.8 Max?
Qwen3.8 Max is the stronger model overall, scoring 56.8 to 38.3 on the Noometry Index. MiniMax-M2.5 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-M2.5 or Qwen3.8 Max?
MiniMax-M2.5 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-M2.5 or Qwen3.8 Max better for coding?
Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 48.1 in the Noometry coding category.
Which has the bigger context window?
Qwen3.8 Max does, with 1M tokens against 205K.
How many benchmarks do MiniMax-M2.5 and Qwen3.8 Max share?
21 benchmarks have published results for both models. MiniMax-M2.5 has 33 scored results on Noometry and Qwen3.8 Max has 39.