# Qwen2-72B vs Qwen3.8 Max

> Qwen3.8 Max is the stronger model overall, scoring 56.8 to 30.0 on the Noometry Index.

- Canonical page: https://noometry.com/compare/qwen2-72b-vs-qwen3-8-max
- Last updated: 2026-10-12
- Shared benchmarks: 19

## Summary

- They share 19 benchmarks with published results for both. Qwen2-72B 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 30.2.
- The biggest single-benchmark swing is GPQA Diamond: 40.8% for Qwen2-72B and 92.7% for Qwen3.8 Max.
- Qwen2-72B has downloadable open weights; the other is API-only.

## Snapshot

| | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 30.0 | 56.8 |
| Rank | 300 | 22 |
| Context | — | 1M |
| Input $/M | — | $2 |
| Output $/M | — | $6 |
| Weights | Open | Proprietary |

## Coding

- Qwen2-72B: 29.1 (#310)
- Qwen3.8 Max: 53.5 (#29)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Coding | 1196 | 1502 |
| DeepSWE | — | 57.5% |
| LMArena WebDev | — | 1674 |
| FrontierSWE | — | 17.8% |
| SciCode | — | 53.2% |
| WeirdML | 11.3% | — |
| BigCodeBench Instruct | 38.5% | — |
| BigCodeBench Complete | 54% | — |

## Agentic & Tool Use

- Qwen2-72B: 17.0 (#146)
- Qwen3.8 Max: 45.4 (#14)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| APEX-Agents | — | 63.3% |
| TheAgentCompany | 1.1% | — |
| τ²-bench Banking | — | 55.1% |
| GDP.pdf | — | 23.2% |
| METR Time Horizons | 29.9% | — |

## Reasoning

- Qwen2-72B: 23.2 (#181)
- Qwen3.8 Max: 54.4 (#26)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Hard Prompts | 1191 | 1496 |
| Epoch Capabilities Index | 125.28 | 156.41 |
| NYT Connections (extended) | — | 88.3% |
| CritPt | — | 20% |
| Chess Puzzles | — | 40% |
| Mystery Game Puzzles | — | 38% |
| DTBench | — | 92% |
| LMCA | — | 46.2% |

## Math

- Qwen2-72B: 30.2 (#236)
- Qwen3.8 Max: 73.2 (#20)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Math | 1235 | 1499 |
| FrontierMath (Tiers 1-3) | — | 74.7% |
| FrontierMath Tier 4 | — | 46.3% |
| OTIS Mock AIME 2024-2025 | — | 100% |
| ProofBench | — | 58% |
| MATH Level 5 | 39.1% | — |

## Knowledge

- Qwen2-72B: 21.2 (#275)
- Qwen3.8 Max: 61.7 (#27)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| GPQA Diamond | 40.8% | 92.7% |
| LMArena Expert | 1171 | 1507 |
| SimpleQA Verified | — | 47.3% |
| MMLU | 82.4% | — |

## Multimodal

- Qwen2-72B: —
- Qwen3.8 Max: 37.2 (#75)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Vision | — | 1314 |
| Furniture Assembly | — | 20% |

## Multilingual

- Qwen2-72B: 35.9 (#244)
- Qwen3.8 Max: 56.7 (#18)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Non-English | 1176 | 1472 |
| LMArena Chinese | 1240 | 1538 |
| LMArena French | 1170 | 1503 |
| LMArena German | 1151 | 1483 |
| LMArena Japanese | 1111 | 1467 |
| LMArena Korean | 1083 | 1461 |
| LMArena Russian | 1169 | 1481 |
| LMArena Spanish | 1169 | 1492 |

## Instruction Following

- Qwen2-72B: 61.7 (#241)
- Qwen3.8 Max: 77.6 (#17)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Instruction Following | 1181 | 1479 |

## Long Context

- Qwen2-72B: 36.1 (#235)
- Qwen3.8 Max: 45.6 (#31)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Longer Query | 1192 | 1489 |

## Writing & Preference

- Qwen2-72B: 40.8 (#241)
- Qwen3.8 Max: 67.1 (#30)

| Benchmark | Qwen2-72B | Qwen3.8 Max |
|---|---|---|
| LMArena Text | 1203 | 1483 |
| LMArena Creative Writing | 1181 | 1479 |
| LMArena Multi-Turn | 1196 | 1489 |

## FAQ

### Is Qwen2-72B better than Qwen3.8 Max?

Qwen3.8 Max is the stronger model overall, scoring 56.8 to 30.0 on the Noometry Index.

### Is Qwen2-72B or Qwen3.8 Max better for coding?

Qwen3.8 Max scores higher on coding benchmarks: 53.5 versus 29.1 in the Noometry coding category.

### How many benchmarks do Qwen2-72B and Qwen3.8 Max share?

19 benchmarks have published results for both models. Qwen2-72B has 26 scored results on Noometry and Qwen3.8 Max has 39.