Model comparison
Qwen2.5 72B Instruct vs Qwen3 Max
Qwen3 Max is the stronger model overall, scoring 43.5 to 31.5 on the Noometry Index.
Last verified . 23 shared benchmarks.
Alibaba (Qwen)Summary
- They share 23 benchmarks with published results for both. Qwen2.5 72B Instruct scores higher in 0 categories and Qwen3 Max in 8 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen3 Max leads 48.1 to 26.9.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 8.1% for Qwen2.5 72B Instruct and 73.3% for Qwen3 Max.
- Both cost about the same: $1.40 input and $5.60 output per million tokens.
- Qwen3 Max accepts more context: 262K tokens versus 131K.
- Qwen2.5 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| Qwen2.5 72B Instruct | Qwen3 Max | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 31.5 | 43.5 |
| Released | 2024-09 | 2025-09-23 |
| Weights | Open | Proprietary |
| Context window | 131K | 262K |
| Max output | 8K | 66K |
| Input $ / M tokens | $1.40 | $1.20 |
| Output $ / M tokens | $5.60 | $6 |
| Results tracked | 43 | 33 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3 Max leads
Qwen2.5 72B Instruct: 32.8 (#256), Qwen3 Max: 42.4 (#93)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Coding | 1293 | 1456 |
| WeirdML | 16% | — |
| BigCodeBench Instruct | 45.8% | — |
| BigCodeBench Complete | 55.9% | — |
| ALE-Bench | — | 370.45 |
Agentic & Tool Use Not comparable
Qwen2.5 72B Instruct: 22.0 (#124), Qwen3 Max: —
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| TheAgentCompany | 5.7% | — |
| BALROG | 16.2% | — |
| METR Time Horizons | 35.8% | — |
| Vending-Bench 2 | — | 71.56 |
Reasoning Too close to call
Qwen2.5 72B Instruct: 21.4 (#197), Qwen3 Max: 22.0 (#186)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Hard Prompts | 1271 | 1448 |
| DTBench | 62.9% | 82.1% |
| LMCA | 13.4% | 28.3% |
| Epoch Capabilities Index | 129.01 | 142.41 |
| Kagi LLM Benchmark | — | 72.5% |
| NYT Connections (extended) | — | 30.1% |
| Chess Puzzles | — | 4% |
| Mystery Game Puzzles | — | 5% |
| BIG-Bench Hard | 79.8% | — |
| ForecastBench | 57.5 | — |
| HellaSwag | 84.8% | — |
| PIQA | 82.6% | — |
| WinoGrande | 82.3% | — |
Math Qwen3 Max leads
Qwen2.5 72B Instruct: 18.1 (#281), Qwen3 Max: 37.6 (#94)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 8.1% | 73.3% |
| LMArena Math | 1283 | 1447 |
| MATH Level 5 | 63.2% | 97.1% |
| FrontierMath (Tiers 1-3) | — | 18.9% |
| Omni-MATH | 33% | — |
Knowledge Qwen3 Max leads
Qwen2.5 72B Instruct: 26.9 (#247), Qwen3 Max: 48.1 (#72)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| GPQA Diamond | 49.1% | 72.6% |
| LMArena Expert | 1245 | 1457 |
| SimpleQA Verified | — | 48.7% |
| MMLU-Pro | 63.1% | — |
| Confabulations | 19.1% | — |
| GPQA (HELM) | 42.6% | — |
| ARC (AI2) Challenge | 94.5% | — |
| MMLU | 85.3% | — |
| TriviaQA | 71.9% | — |
Multilingual Qwen3 Max leads
Qwen2.5 72B Instruct: 41.2 (#208), Qwen3 Max: 53.9 (#58)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Non-English | 1252 | 1429 |
| LMArena Chinese | 1272 | 1479 |
| LMArena French | 1279 | 1450 |
| LMArena German | 1234 | 1463 |
| LMArena Japanese | 1180 | 1399 |
| LMArena Korean | 1188 | 1397 |
| LMArena Russian | 1264 | 1428 |
| LMArena Spanish | 1255 | 1463 |
Instruction Following Qwen3 Max leads
Qwen2.5 72B Instruct: 65.7 (#216), Qwen3 Max: 75.0 (#81)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Instruction Following | 1254 | 1420 |
| IFEval | 80.6% | — |
Long Context Qwen3 Max leads
Qwen2.5 72B Instruct: 38.9 (#183), Qwen3 Max: 41.6 (#129)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Longer Query | 1282 | 1439 |
| Fiction.LiveBench | — | 66.7% |
| CL-bench | — | 14.5% |
Writing & Preference Qwen3 Max leads
Qwen2.5 72B Instruct: 46.9 (#210), Qwen3 Max: 62.6 (#72)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 Max |
|---|---|---|
| LMArena Text | 1269 | 1439 |
| LMArena Creative Writing | 1221 | 1401 |
| LMArena Multi-Turn | 1272 | 1446 |
| WildBench | 80.2% | — |
Frequently asked questions
Is Qwen2.5 72B Instruct better than Qwen3 Max?
Qwen3 Max is the stronger model overall, scoring 43.5 to 31.5 on the Noometry Index.
Which is cheaper, Qwen2.5 72B Instruct or Qwen3 Max?
Qwen3 Max is cheaper. It lists at $1.20 per million input tokens and $6 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is Qwen2.5 72B Instruct or Qwen3 Max better for coding?
Qwen3 Max scores higher on coding benchmarks: 42.4 versus 32.8 in the Noometry coding category.
Which has the bigger context window?
Qwen3 Max does, with 262K tokens against 131K.
How many benchmarks do Qwen2.5 72B Instruct and Qwen3 Max share?
23 benchmarks have published results for both models. Qwen2.5 72B Instruct has 43 scored results on Noometry and Qwen3 Max has 33.