Model comparison
Qwen2.5 72B Instruct vs Qwen3 32B
Qwen3 32B is the stronger model overall, scoring 38.8 to 31.5 on the Noometry Index.
Last verified . 18 shared benchmarks.
Alibaba (Qwen)Summary
- They share 18 benchmarks with published results for both. Qwen2.5 72B Instruct scores higher in 1 category and Qwen3 32B in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where Qwen3 32B leads 37.9 to 18.1.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 8.1% for Qwen2.5 72B Instruct and 66.9% for Qwen3 32B.
- Qwen3 32B is cheaper at $0.70 / $2.80 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
Side by side
| Qwen2.5 72B Instruct | Qwen3 32B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 31.5 | 38.8 |
| Released | 2024-09 | 2025-04 |
| Weights | Open | Open |
| Context window | 131K | 131K |
| Max output | 8K | 16K |
| Input $ / M tokens | $1.40 | $0.70 |
| Output $ / M tokens | $5.60 | $2.80 |
| Results tracked | 43 | 26 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen3 32B leads
Qwen2.5 72B Instruct: 32.8 (#256), Qwen3 32B: 37.3 (#184)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Coding | 1293 | 1358 |
| Aider Polyglot | — | 40% |
| SciCode | — | 35.4% |
| WeirdML | 16% | — |
| BigCodeBench Instruct | 45.8% | — |
| BigCodeBench Complete | 55.9% | — |
Agentic & Tool Use Qwen3 32B leads
Qwen2.5 72B Instruct: 22.0 (#124), Qwen3 32B: 32.3 (#58)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 48.7% |
| TheAgentCompany | 5.7% | — |
| BALROG | 16.2% | — |
| METR Time Horizons | 35.8% | — |
Reasoning Qwen2.5 72B Instruct leads
Qwen2.5 72B Instruct: 21.4 (#197), Qwen3 32B: 19.6 (#233)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Hard Prompts | 1271 | 1335 |
| DTBench | 62.9% | 67.5% |
| LMCA | 13.4% | 17.3% |
| Epoch Capabilities Index | 129.01 | 138.51 |
| Kagi LLM Benchmark | — | 54.9% |
| CritPt | — | 0.3% |
| Chess Puzzles | — | 5% |
| BIG-Bench Hard | 79.8% | — |
| ForecastBench | 57.5 | — |
| HellaSwag | 84.8% | — |
| PIQA | 82.6% | — |
| WinoGrande | 82.3% | — |
Math Qwen3 32B leads
Qwen2.5 72B Instruct: 18.1 (#281), Qwen3 32B: 37.9 (#89)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 8.1% | 66.9% |
| LMArena Math | 1283 | 1399 |
| Omni-MATH | 33% | — |
| MATH Level 5 | 63.2% | — |
Knowledge Qwen3 32B leads
Qwen2.5 72B Instruct: 26.9 (#247), Qwen3 32B: 40.0 (#112)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| GPQA Diamond | 49.1% | 65.7% |
| LMArena Expert | 1245 | 1362 |
| MMLU-Pro | 63.1% | — |
| Confabulations | 19.1% | — |
| Vectara Hallucination Rate | — | 5.9% |
| GPQA (HELM) | 42.6% | — |
| ARC (AI2) Challenge | 94.5% | — |
| MMLU | 85.3% | — |
| TriviaQA | 71.9% | — |
Multilingual Qwen3 32B leads
Qwen2.5 72B Instruct: 41.2 (#208), Qwen3 32B: 45.8 (#162)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Non-English | 1252 | 1317 |
| LMArena Chinese | 1272 | 1357 |
| LMArena German | 1234 | 1341 |
| LMArena Russian | 1264 | 1311 |
| LMArena French | 1279 | — |
| LMArena Japanese | 1180 | — |
| LMArena Korean | 1188 | — |
| LMArena Spanish | 1255 | — |
Instruction Following Qwen3 32B leads
Qwen2.5 72B Instruct: 65.7 (#216), Qwen3 32B: 69.1 (#174)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Instruction Following | 1254 | 1305 |
| IFEval | 80.6% | — |
Long Context Qwen3 32B leads
Qwen2.5 72B Instruct: 38.9 (#183), Qwen3 32B: 43.7 (#83)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Longer Query | 1282 | 1327 |
| Fiction.LiveBench | — | 74.2% |
Writing & Preference Qwen3 32B leads
Qwen2.5 72B Instruct: 46.9 (#210), Qwen3 32B: 53.1 (#158)
| Benchmark | Qwen2.5 72B Instruct | Qwen3 32B |
|---|---|---|
| LMArena Text | 1269 | 1340 |
| LMArena Creative Writing | 1221 | 1297 |
| LMArena Multi-Turn | 1272 | 1331 |
| WildBench | 80.2% | — |
Frequently asked questions
Is Qwen2.5 72B Instruct better than Qwen3 32B?
Qwen3 32B is the stronger model overall, scoring 38.8 to 31.5 on the Noometry Index.
Which is cheaper, Qwen2.5 72B Instruct or Qwen3 32B?
Qwen3 32B is cheaper. It lists at $0.70 per million input tokens and $2.80 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is Qwen2.5 72B Instruct or Qwen3 32B better for coding?
Qwen3 32B scores higher on coding benchmarks: 37.3 versus 32.8 in the Noometry coding category.
Which has the bigger context window?
Both accept 131K tokens.
How many benchmarks do Qwen2.5 72B Instruct and Qwen3 32B share?
18 benchmarks have published results for both models. Qwen2.5 72B Instruct has 43 scored results on Noometry and Qwen3 32B has 26.