Model comparison
GPT-4o vs Qwen2.5 72B Instruct
Qwen2.5 72B Instruct is the stronger model overall, scoring 31.5 to 28.5 on the Noometry Index.
Last verified . 37 shared benchmarks.
Alibaba (Qwen)Summary
- They share 37 benchmarks with published results for both. GPT-4o scores higher in 5 categories and Qwen2.5 72B Instruct in 4 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Qwen2.5 72B Instruct leads 21.4 to 9.4.
- The biggest single-benchmark swing is BALROG: 32.3% for GPT-4o and 16.2% for Qwen2.5 72B Instruct.
- Qwen2.5 72B Instruct is cheaper at $1.40 / $5.60 per million input/output tokens, against $2.50 / $10 for GPT-4o.
- Qwen2.5 72B Instruct accepts more context: 131K tokens versus 128K.
- Qwen2.5 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| GPT-4o | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 28.5 | 31.5 |
| Released | 2024-05-13 | 2024-09 |
| Weights | Proprietary | Open |
| Context window | 128K | 131K |
| Max output | 16K | 8K |
| Input $ / M tokens | $2.50 | $1.40 |
| Output $ / M tokens | $10 | $5.60 |
| Results tracked | 72 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Qwen2.5 72B Instruct leads
GPT-4o: 25.1 (#323), Qwen2.5 72B Instruct: 32.8 (#256)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 25.1% | 16% |
| BigCodeBench Instruct | 51.1% | 45.8% |
| LMArena Coding | 1297 | 1293 |
| BigCodeBench Complete | 61.1% | 55.9% |
| SWE-bench Verified | 31% | — |
| SWE-bench Verified (bash only) | 21.6% | — |
| Aider Polyglot | 45.3% | — |
| GSO | 0% | — |
| LiveBench Coding | 51.4% | — |
| CadEval | 26% | — |
| HumanEval+ | 87.2% | — |
| MBPP+ | 72.2% | — |
Agentic & Tool Use Qwen2.5 72B Instruct leads
GPT-4o: 20.8 (#133), Qwen2.5 72B Instruct: 22.0 (#124)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| TheAgentCompany | 8.6% | 5.7% |
| BALROG | 32.3% | 16.2% |
| METR Time Horizons | 40.8% | 35.8% |
| GDPval | 9.9% | — |
| Cybench | 12.5% | — |
| LMArena Search | 1006 | — |
Reasoning Qwen2.5 72B Instruct leads
GPT-4o: 9.4 (#338), Qwen2.5 72B Instruct: 21.4 (#197)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1281 | 1271 |
| DTBench | 64.5% | 62.9% |
| LMCA | 16.6% | 13.4% |
| Epoch Capabilities Index | 128.97 | 129.01 |
| ForecastBench | 57.7 | 57.5 |
| ARC-AGI-2 | 0% | — |
| SimpleBench | 17.8% | — |
| ARC-AGI-1 | 4.5% | — |
| CritPt | 0% | — |
| Chess Puzzles | 13% | — |
| EnigmaEval | 0.8% | — |
| LiveBench Reasoning | 55.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | — | 79.8% |
| HellaSwag | — | 84.8% |
| LiveBench | 55.3% | — |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math Qwen2.5 72B Instruct leads
GPT-4o: 10.2 (#303), Qwen2.5 72B Instruct: 18.1 (#281)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 6.4% | 8.1% |
| Omni-MATH | 29.3% | 33% |
| LMArena Math | 1285 | 1283 |
| MATH Level 5 | 53.3% | 63.2% |
| FrontierMath (Tiers 1-3) | 0.4% | — |
| LiveBench Math | 49.5% | — |
| FrontierMath (Feb 2025 set) | 0.3% | — |
Knowledge GPT-4o leads
GPT-4o: 28.8 (#235), Qwen2.5 72B Instruct: 26.9 (#247)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 49.2% | 49.1% |
| MMLU-Pro | 71.3% | 63.1% |
| Confabulations | 15.3% | 19.1% |
| GPQA (HELM) | 52% | 42.6% |
| LMArena Expert | 1250 | 1245 |
| MMLU | 88.1% | 85.3% |
| Humanity's Last Exam | 2.7% | — |
| SimpleQA Verified | 26% | — |
| Vectara Hallucination Rate | 9.6% | — |
| ARC (AI2) Challenge | — | 94.5% |
| TriviaQA | — | 71.9% |
Multimodal Not comparable
GPT-4o: 33.9 (#85), Qwen2.5 72B Instruct: —
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Vision | 1137 | — |
| Video-MME | 71.9% | — |
| GeoBench | 71% | — |
| VPCT | 40% | — |
| ScienceQA | 88.5% | — |
Multilingual GPT-4o leads
GPT-4o: 43.4 (#181), Qwen2.5 72B Instruct: 41.2 (#208)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1283 | 1252 |
| LMArena Chinese | 1277 | 1272 |
| LMArena French | 1303 | 1279 |
| LMArena German | 1282 | 1234 |
| LMArena Japanese | 1257 | 1180 |
| LMArena Korean | 1234 | 1188 |
| LMArena Russian | 1286 | 1264 |
| LMArena Spanish | 1292 | 1255 |
Instruction Following GPT-4o leads
GPT-4o: 66.8 (#202), Qwen2.5 72B Instruct: 65.7 (#216)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| IFEval | 81.7% | 80.6% |
| LMArena Instruction Following | 1278 | 1254 |
| LiveBench Instruction Following | 68.6% | — |
Long Context Too close to call
GPT-4o: 39.4 (#174), Qwen2.5 72B Instruct: 38.9 (#183)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1289 | 1282 |
| Fiction.LiveBench | 66.7% | — |
Writing & Preference GPT-4o leads
GPT-4o: 52.8 (#160), Qwen2.5 72B Instruct: 46.9 (#210)
| Benchmark | GPT-4o | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1300 | 1269 |
| LMArena Creative Writing | 1291 | 1221 |
| WildBench | 82.8% | 80.2% |
| LMArena Multi-Turn | 1302 | 1272 |
| Short-Story Creative Writing | 81.8% | — |
| LiveBench Language | 47.6% | — |
Frequently asked questions
Is GPT-4o better than Qwen2.5 72B Instruct?
Qwen2.5 72B Instruct is the stronger model overall, scoring 31.5 to 28.5 on the Noometry Index.
Which is cheaper, GPT-4o or Qwen2.5 72B Instruct?
Qwen2.5 72B Instruct is cheaper. It lists at $1.40 per million input tokens and $5.60 per million output tokens; GPT-4o lists at $2.50 and $10.
Is GPT-4o or Qwen2.5 72B Instruct better for coding?
Qwen2.5 72B Instruct scores higher on coding benchmarks: 32.8 versus 25.1 in the Noometry coding category.
Which has the bigger context window?
Qwen2.5 72B Instruct does, with 131K tokens against 128K.
How many benchmarks do GPT-4o and Qwen2.5 72B Instruct share?
37 benchmarks have published results for both models. GPT-4o has 72 scored results on Noometry and Qwen2.5 72B Instruct has 43.