Model comparison
Qwen2-72B vs Qwen3-4B
Qwen3-4B is the stronger model overall, scoring 31.9 to 30.0 on the Noometry Index.
Last verified . 1 shared benchmarks.
Alibaba (Qwen)Summary
- They share 1 benchmark with published results for both. Qwen2-72B scores higher in 2 categories and Qwen3-4B in 2 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where Qwen3-4B leads 33.0 to 21.2.
- The biggest single-benchmark swing is GPQA Diamond: 40.8% for Qwen2-72B and 52.3% for Qwen3-4B.
Side by side
| Qwen2-72B | Qwen3-4B | |
|---|---|---|
| Provider | Alibaba (Qwen) | Alibaba (Qwen) |
| Noometry Index | 30.0 | 31.9 |
| Released | 2024-06-07 | 2025-04-29 |
| Weights | Open | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 26 | 6 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
Qwen2-72B: 29.1 (#310), Qwen3-4B: —
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| WeirdML | 11.3% | — |
| BigCodeBench Instruct | 38.5% | — |
| LMArena Coding | 1196 | — |
| BigCodeBench Complete | 54% | — |
Agentic & Tool Use Qwen3-4B leads
Qwen2-72B: 17.0 (#146), Qwen3-4B: 27.6 (#100)
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 35.7% |
| TheAgentCompany | 1.1% | — |
| METR Time Horizons | 29.9% | — |
Reasoning Qwen2-72B leads
Qwen2-72B: 23.2 (#181), Qwen3-4B: 19.2 (#268)
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| Chess Puzzles | — | 4% |
| LMArena Hard Prompts | 1191 | — |
| Epoch Capabilities Index | 125.28 | — |
Math Too close to call
Qwen2-72B: 30.2 (#236), Qwen3-4B: 29.7 (#240)
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| MathArena Final-Answer Competitions | — | 38.5% |
| OTIS Mock AIME 2024-2025 | — | 52.2% |
| LMArena Math | 1235 | — |
| MATH Level 5 | 39.1% | — |
Knowledge Qwen3-4B leads
Qwen2-72B: 21.2 (#275), Qwen3-4B: 33.0 (#208)
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| GPQA Diamond | 40.8% | 52.3% |
| Vectara Hallucination Rate | — | 5.7% |
| LMArena Expert | 1171 | — |
| MMLU | 82.4% | — |
Multilingual Not comparable
Qwen2-72B: 35.9 (#244), Qwen3-4B: —
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| LMArena Non-English | 1176 | — |
| LMArena Chinese | 1240 | — |
| LMArena French | 1170 | — |
| LMArena German | 1151 | — |
| LMArena Japanese | 1111 | — |
| LMArena Korean | 1083 | — |
| LMArena Russian | 1169 | — |
| LMArena Spanish | 1169 | — |
Instruction Following Not comparable
Qwen2-72B: 61.7 (#241), Qwen3-4B: —
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| LMArena Instruction Following | 1181 | — |
Long Context Not comparable
Qwen2-72B: 36.1 (#235), Qwen3-4B: —
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| LMArena Longer Query | 1192 | — |
Writing & Preference Not comparable
Qwen2-72B: 40.8 (#241), Qwen3-4B: —
| Benchmark | Qwen2-72B | Qwen3-4B |
|---|---|---|
| LMArena Text | 1203 | — |
| LMArena Creative Writing | 1181 | — |
| LMArena Multi-Turn | 1196 | — |
Frequently asked questions
Is Qwen2-72B better than Qwen3-4B?
Qwen3-4B is the stronger model overall, scoring 31.9 to 30.0 on the Noometry Index.
How many benchmarks do Qwen2-72B and Qwen3-4B share?
1 benchmark has published results for both models. Qwen2-72B has 26 scored results on Noometry and Qwen3-4B has 6.