Model comparison
Qwen2-72B vs Qwen3.5 27B
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 30.0 on the Noometry Index.
Last verified . 18 shared benchmarks.
Alibaba (Qwen)Summary
- They share 18 benchmarks with published results for both. Qwen2-72B scores higher in 0 categories and Qwen3.5 27B in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Qwen3.5 27B leads 59.3 to 40.8.
- The biggest single-benchmark swing is WeirdML: 11.3% for Qwen2-72B and 39.5% for Qwen3.5 27B.
Side by side
| Qwen2-72B | Qwen3.5 27B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 30.0 | 41.9 |
| Released | 2024-06-07 | 2026-02-23 |
| Weights | Open | Open |
| Context window | — | 262K |
| Max output | — | 66K |
| Input $ / M tokens | — | $0.30 |
| Output $ / M tokens | — | $2.40 |
| Results tracked | 26 | 28 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3.5 27B leads
Qwen2-72B: 29.1 (#310), Qwen3.5 27B: 38.9 (#168)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| WeirdML | 11.3% | 39.5% |
| LMArena Coding | 1196 | 1427 |
| LMArena WebDev | — | 1358 |
| BigCodeBench Instruct | 38.5% | — |
| BigCodeBench Complete | 54% | — |
| ALE-Bench | — | 349.45 |
Agentic & Tool Use Not comparable
Qwen2-72B: 17.0 (#146), Qwen3.5 27B: —
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| TheAgentCompany | 1.1% | — |
| METR Time Horizons | 29.9% | — |
| Vending-Bench 2 | — | 201.98 |
Reasoning Qwen3.5 27B leads
Qwen2-72B: 23.2 (#181), Qwen3.5 27B: 27.5 (#117)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Hard Prompts | 1191 | 1414 |
| NYT Connections (extended) | — | 47.9% |
| Thematic Generalization | — | 45.5% |
| DTBench | — | 82.4% |
| LMCA | — | 34% |
| Epoch Capabilities Index | 125.28 | — |
Math Qwen3.5 27B leads
Qwen2-72B: 30.2 (#236), Qwen3.5 27B: 38.8 (#127)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Math | 1235 | 1429 |
| MathArena Final-Answer Competitions | — | 56.7% |
| MATH Level 5 | 39.1% | — |
Knowledge Qwen3.5 27B leads
Qwen2-72B: 21.2 (#275), Qwen3.5 27B: 38.0 (#150)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Expert | 1171 | 1428 |
| GPQA Diamond | 40.8% | — |
| Vectara Hallucination Rate | — | 12.1% |
| MMLU | 82.4% | — |
Multimodal Not comparable
Qwen2-72B: —, Qwen3.5 27B: 39.4 (#59)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Vision | — | 1241 |
Multilingual Qwen3.5 27B leads
Qwen2-72B: 35.9 (#244), Qwen3.5 27B: 50.8 (#115)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Non-English | 1176 | 1390 |
| LMArena Chinese | 1240 | 1478 |
| LMArena French | 1170 | 1410 |
| LMArena German | 1151 | 1393 |
| LMArena Japanese | 1111 | 1345 |
| LMArena Korean | 1083 | 1358 |
| LMArena Russian | 1169 | 1390 |
| LMArena Spanish | 1169 | 1407 |
Instruction Following Qwen3.5 27B leads
Qwen2-72B: 61.7 (#241), Qwen3.5 27B: 73.5 (#119)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Instruction Following | 1181 | 1393 |
Long Context Qwen3.5 27B leads
Qwen2-72B: 36.1 (#235), Qwen3.5 27B: 43.1 (#106)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Longer Query | 1192 | 1413 |
Writing & Preference Qwen3.5 27B leads
Qwen2-72B: 40.8 (#241), Qwen3.5 27B: 59.3 (#111)
| Benchmark | Qwen2-72B | Qwen3.5 27B |
|---|---|---|
| LMArena Text | 1203 | 1409 |
| LMArena Creative Writing | 1181 | 1362 |
| LMArena Multi-Turn | 1196 | 1410 |
Frequently asked questions
Is Qwen2-72B better than Qwen3.5 27B?
Qwen3.5 27B is the stronger model overall, scoring 41.9 to 30.0 on the Noometry Index.
Is Qwen2-72B or Qwen3.5 27B better for coding?
Qwen3.5 27B scores higher on coding benchmarks: 38.9 versus 29.1 in the Noometry coding category.
How many benchmarks do Qwen2-72B and Qwen3.5 27B share?
18 benchmarks have published results for both models. Qwen2-72B has 26 scored results on Noometry and Qwen3.5 27B has 28.