Model comparison
Qwen1.5-32B vs Qwen1.5-72B
Qwen1.5-32B and Qwen1.5-72B score almost the same on the Noometry Index (30.5 vs 30.8), so choose on price, context window or the category you care about most.
Last verified . 20 shared benchmarks.
Alibaba (Qwen)Summary
- They share 20 benchmarks with published results for both. Qwen1.5-32B scores higher in 1 category and Qwen1.5-72B in 7 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Qwen1.5-72B leads 37.3 to 34.2.
Side by side
| Qwen1.5-32B | Qwen1.5-72B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 30.5 | 30.8 |
| Released | 2024-02-04 | 2024-02-04 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 21 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Qwen1.5-32B: 31.7 (#282), Qwen1.5-72B: 31.9 (#277)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| BigCodeBench Instruct | 32.3% | 33.2% |
| LMArena Coding | 1155 | 1165 |
| BigCodeBench Complete | 42% | 40.3% |
| HumanEval+ | — | 59.1% |
| MBPP+ | — | 61.6% |
Reasoning Too close to call
Qwen1.5-32B: 21.8 (#212), Qwen1.5-72B: 22.2 (#203)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Hard Prompts | 1130 | 1148 |
Math Too close to call
Qwen1.5-32B: 33.0 (#207), Qwen1.5-72B: 33.2 (#205)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Math | 1155 | 1164 |
Knowledge Qwen1.5-32B leads
Qwen1.5-32B: 13.5 (#296), Qwen1.5-72B: 11.5 (#300)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| GPQA Diamond | 30.7% | 28.8% |
| LMArena Expert | 1126 | 1136 |
| MMLU | 74.4% | — |
Multilingual Qwen1.5-72B leads
Qwen1.5-32B: 31.4 (#259), Qwen1.5-72B: 33.2 (#253)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Non-English | 1106 | 1135 |
| LMArena Chinese | 1177 | 1186 |
| LMArena French | 1101 | 1159 |
| LMArena German | 1058 | 1084 |
| LMArena Japanese | 1027 | 1061 |
| LMArena Korean | 1008 | 1050 |
| LMArena Russian | 1073 | 1104 |
| LMArena Spanish | 1089 | 1110 |
Instruction Following Qwen1.5-72B leads
Qwen1.5-32B: 57.7 (#265), Qwen1.5-72B: 59.3 (#256)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Instruction Following | 1116 | 1141 |
Long Context Too close to call
Qwen1.5-32B: 34.7 (#246), Qwen1.5-72B: 35.1 (#243)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Longer Query | 1146 | 1157 |
Writing & Preference Qwen1.5-72B leads
Qwen1.5-32B: 34.2 (#271), Qwen1.5-72B: 37.3 (#258)
| Benchmark | Qwen1.5-32B | Qwen1.5-72B |
|---|---|---|
| LMArena Text | 1137 | 1166 |
| LMArena Creative Writing | 1083 | 1137 |
| LMArena Multi-Turn | 1140 | 1160 |
Frequently asked questions
Is Qwen1.5-32B better than Qwen1.5-72B?
Qwen1.5-32B and Qwen1.5-72B score almost the same on the Noometry Index (30.5 vs 30.8), so choose on price, context window or the category you care about most.
Is Qwen1.5-32B or Qwen1.5-72B better for coding?
They score almost the same on coding (31.7 vs 31.9); test both on your own repository before choosing.
How many benchmarks do Qwen1.5-32B and Qwen1.5-72B share?
20 benchmarks have published results for both models. Qwen1.5-32B has 21 scored results on Noometry and Qwen1.5-72B has 22.