Model comparison
MiniMax-M3 vs Qwen3.5 397B-A17B
Qwen3.5 397B-A17B is the stronger model overall, scoring 46.0 to 43.8 on the Noometry Index. MiniMax-M3 costs 2.6× less per token, which makes it the better buy when Qwen3.5 397B-A17B's lead doesn't matter for your workload.
Last verified . 28 shared benchmarks.
MiniMax
Alibaba (Qwen)Summary
- They share 28 benchmarks with published results for both. MiniMax-M3 scores higher in 3 categories and Qwen3.5 397B-A17B in 7 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in agentic & tool use, where Qwen3.5 397B-A17B leads 33.3 to 22.6.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 71.1% for MiniMax-M3 and 88.9% for Qwen3.5 397B-A17B.
- MiniMax-M3 is cheaper at $0.30 / $1.20 per million input/output tokens, against $0.60 / $3.60 for Qwen3.5 397B-A17B.
- MiniMax-M3 accepts more context: 1M tokens versus 262K.
Side by side
| MiniMax-M3 | Qwen3.5 397B-A17B | |
|---|---|---|
| Provider | MiniMax | Alibaba (Qwen) |
| Noometry Index | 43.8 | 46.0 |
| Released | 2026-06-01 | 2026-02-01 |
| Weights | Open | Open |
| Context window | 1M | 262K |
| Max output | 512K | 66K |
| Input $ / M tokens | $0.30 | $0.60 |
| Output $ / M tokens | $1.20 | $3.60 |
| Results tracked | 41 | 36 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
MiniMax-M3: 41.8 (#118), Qwen3.5 397B-A17B: 42.0 (#114)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena WebDev | 1482 | 1400 |
| LMArena Coding | 1469 | 1465 |
| FrontierCode | 14.7% | — |
| SciCode | 47.1% | — |
| ALE-Bench | 640.02 | — |
Agentic & Tool Use Qwen3.5 397B-A17B leads
MiniMax-M3: 22.6 (#130), Qwen3.5 397B-A17B: 33.3 (#53)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| APEX-Agents | 37.7% | 24.9% |
| OSWorld 2.0 | 4.6% | — |
| τ²-bench Airline | — | 81.5% |
| τ²-bench Banking | — | 9.8% |
| τ²-bench Retail | — | 84.4% |
| τ²-bench Telecom | — | 97.8% |
| GBAEval | 0.9% | — |
| Vending-Bench 2 | 2,158 | — |
Reasoning Qwen3.5 397B-A17B leads
MiniMax-M3: 30.1 (#87), Qwen3.5 397B-A17B: 34.5 (#70)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| NYT Connections (extended) | 65.1% | 58.9% |
| Chess Puzzles | 14% | 13% |
| LMArena Hard Prompts | 1447 | 1448 |
| Mystery Game Puzzles | 8% | 18% |
| DTBench | 78.9% | 87.5% |
| LMCA | 33.7% | 37.9% |
| Epoch Capabilities Index | 146.95 | 146.65 |
| SimpleBench | 45.8% | — |
| Kagi LLM Benchmark | — | 73.7% |
| CritPt | 3.7% | — |
| Thematic Generalization | — | 65.1% |
| Surface Evolver Bench | 55% | — |
| ForecastBench | 61.4 | — |
Math Qwen3.5 397B-A17B leads
MiniMax-M3: 40.0 (#95), Qwen3.5 397B-A17B: 46.1 (#73)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 71.1% | 88.9% |
| LMArena Math | 1429 | 1454 |
| FrontierMath (Tiers 1-3) | — | 31.2% |
| ProofBench | 18% | — |
Knowledge MiniMax-M3 leads
MiniMax-M3: 58.4 (#35), Qwen3.5 397B-A17B: 53.3 (#58)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| GPQA Diamond | 90.9% | 86.4% |
| LMArena Expert | 1461 | 1462 |
Multimodal Too close to call
MiniMax-M3: 40.2 (#51), Qwen3.5 397B-A17B: 40.7 (#44)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Vision | 1253 | 1263 |
| LMArena Document | 1435 | — |
Multilingual Too close to call
MiniMax-M3: 53.0 (#75), Qwen3.5 397B-A17B: 53.7 (#59)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Non-English | 1420 | 1430 |
| LMArena Chinese | 1463 | 1500 |
| LMArena French | 1447 | 1461 |
| LMArena German | 1426 | 1447 |
| LMArena Japanese | 1381 | 1426 |
| LMArena Korean | 1372 | 1384 |
| LMArena Russian | 1428 | 1429 |
| LMArena Spanish | 1432 | 1441 |
Instruction Following Too close to call
MiniMax-M3: 75.5 (#62), Qwen3.5 397B-A17B: 75.0 (#77)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Instruction Following | 1433 | 1424 |
Long Context Too close to call
MiniMax-M3: 44.2 (#72), Qwen3.5 397B-A17B: 44.1 (#74)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Longer Query | 1445 | 1442 |
Writing & Preference Too close to call
MiniMax-M3: 62.1 (#83), Qwen3.5 397B-A17B: 62.3 (#79)
| Benchmark | MiniMax-M3 | Qwen3.5 397B-A17B |
|---|---|---|
| LMArena Text | 1433 | 1438 |
| LMArena Creative Writing | 1404 | 1401 |
| LMArena Multi-Turn | 1442 | 1446 |
| EQ-Bench Creative Writing | — | 1478 |
| EQ-Bench 4 | 1150 | — |
Frequently asked questions
Is MiniMax-M3 better than Qwen3.5 397B-A17B?
Qwen3.5 397B-A17B is the stronger model overall, scoring 46.0 to 43.8 on the Noometry Index. MiniMax-M3 costs 2.6× less per token, which makes it the better buy when Qwen3.5 397B-A17B's lead doesn't matter for your workload.
Which is cheaper, MiniMax-M3 or Qwen3.5 397B-A17B?
MiniMax-M3 is cheaper. It lists at $0.30 per million input tokens and $1.20 per million output tokens; Qwen3.5 397B-A17B lists at $0.60 and $3.60.
Is MiniMax-M3 or Qwen3.5 397B-A17B better for coding?
They score almost the same on coding (41.8 vs 42.0); test both on your own repository before choosing.
Which has the bigger context window?
MiniMax-M3 does, with 1M tokens against 262K.
How many benchmarks do MiniMax-M3 and Qwen3.5 397B-A17B share?
28 benchmarks have published results for both models. MiniMax-M3 has 41 scored results on Noometry and Qwen3.5 397B-A17B has 36.