Model comparison
GPT-5.6 Luna vs Qwen2.5 72B Instruct
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 31.9 on the Noometry Index.
Last verified . 24 shared benchmarks.
Alibaba (Qwen)Summary
- They share 24 benchmarks with published results for both. GPT-5.6 Luna scores higher in 9 categories and Qwen2.5 72B Instruct in 0 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.7 to 19.3.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 98.3% for GPT-5.6 Luna and 8.1% for Qwen2.5 72B Instruct.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $1.40 / $5.60 for Qwen2.5 72B Instruct.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 131K.
- Qwen2.5 72B Instruct has downloadable open weights; the other is API-only.
Side by side
| GPT-5.6 Luna | Qwen2.5 72B Instruct | |
|---|---|---|
| Provider | OpenAI | Alibaba (Qwen) |
| Noometry Index | 54.6 | 31.9 |
| Released | 2026-07-09 | 2024-09 |
| Weights | Proprietary | Open |
| Context window | 1.05M | 131K |
| Max output | 128K | 8K |
| Input $ / M tokens | $0.20 | $1.40 |
| Output $ / M tokens | $1.20 | $5.60 |
| Results tracked | 52 | 43 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
GPT-5.6 Luna: 54.5 (#28), Qwen2.5 72B Instruct: 33.2 (#260)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| WeirdML | 60.9% | 16% |
| LMArena Coding | 1466 | 1292 |
| DeepSWE | 67.2% | — |
| FrontierCode | 39.8% | — |
| CursorBench | 35.9% | — |
| LMArena WebDev | 1519 | — |
| SciCode | 53.6% | — |
| BigCodeBench Instruct | — | 45.8% |
| BigCodeBench Complete | — | 55.9% |
| ALE-Bench | 1,667 | — |
Agentic & Tool Use GPT-5.6 Luna leads
GPT-5.6 Luna: 34.4 (#45), Qwen2.5 72B Instruct: 22.1 (#133)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| BALROG | 45.6% | 16.2% |
| APEX-Agents | 43% | — |
| TheAgentCompany | — | 5.7% |
| GDP.pdf | 22.7% | — |
| METR Time Horizons | — | 35.8% |
| Vending-Bench 2 | 4,095 | — |
Reasoning GPT-5.6 Luna leads
GPT-5.6 Luna: 47.6 (#43), Qwen2.5 72B Instruct: 22.3 (#199)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Hard Prompts | 1451 | 1271 |
| DTBench | 89.1% | 62.9% |
| LMCA | 48.5% | 13.4% |
| Epoch Capabilities Index | 156.39 | 129 |
| ARC-AGI-2 | 59.5% | — |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 49.1% | — |
| NYT Connections (extended) | 69.4% | — |
| ARC-AGI-1 | 88% | — |
| CritPt | 20.6% | — |
| Chess Puzzles | 40% | — |
| Mystery Game Puzzles | 21% | — |
| Surface Evolver Bench | 61.9% | — |
| BIG-Bench Hard | — | 79.8% |
| ForecastBench | — | 57.5 |
| HellaSwag | — | 84.8% |
| PIQA | — | 82.6% |
| WinoGrande | — | 82.3% |
Math GPT-5.6 Luna leads
GPT-5.6 Luna: 77.7 (#14), Qwen2.5 72B Instruct: 19.3 (#287)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 98.3% | 8.1% |
| LMArena Math | 1458 | 1283 |
| FrontierMath (Tiers 1-3) | 82.1% | — |
| FrontierMath Tier 4 | 61% | — |
| ProofBench | 60% | — |
| Omni-MATH | — | 33% |
| MATH Level 5 | — | 63.2% |
Knowledge GPT-5.6 Luna leads
GPT-5.6 Luna: 58.5 (#34), Qwen2.5 72B Instruct: 27.0 (#253)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| GPQA Diamond | 91.6% | 49.1% |
| LMArena Expert | 1478 | 1245 |
| SimpleQA Verified | 41% | — |
| MMLU-Pro | — | 63.1% |
| Confabulations | — | 19.1% |
| GPQA (HELM) | — | 42.6% |
| ARC (AI2) Challenge | — | 94.5% |
| MMLU | — | 85.3% |
| TriviaQA | — | 71.9% |
Multimodal Not comparable
GPT-5.6 Luna: 42.7 (#28), Qwen2.5 72B Instruct: —
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Vision | 1258 | — |
| Blueprint-Bench 2 | 22.6% | — |
| Furniture Assembly | 42.5% | — |
| LMArena Document | 1457 | — |
Multilingual GPT-5.6 Luna leads
GPT-5.6 Luna: 52.8 (#78), Qwen2.5 72B Instruct: 41.0 (#213)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Non-English | 1417 | 1252 |
| LMArena Chinese | 1470 | 1272 |
| LMArena French | 1456 | 1280 |
| LMArena German | 1454 | 1234 |
| LMArena Japanese | 1411 | 1180 |
| LMArena Korean | 1415 | 1188 |
| LMArena Russian | 1428 | 1264 |
| LMArena Spanish | 1448 | 1256 |
Instruction Following GPT-5.6 Luna leads
GPT-5.6 Luna: 75.6 (#57), Qwen2.5 72B Instruct: 65.5 (#221)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Instruction Following | 1437 | 1254 |
| IFEval | — | 80.6% |
Long Context GPT-5.6 Luna leads
GPT-5.6 Luna: 43.9 (#82), Qwen2.5 72B Instruct: 38.9 (#188)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Longer Query | 1436 | 1282 |
Writing & Preference GPT-5.6 Luna leads
GPT-5.6 Luna: 68.0 (#29), Qwen2.5 72B Instruct: 46.7 (#215)
| Benchmark | GPT-5.6 Luna | Qwen2.5 72B Instruct |
|---|---|---|
| LMArena Text | 1431 | 1269 |
| LMArena Creative Writing | 1396 | 1221 |
| LMArena Multi-Turn | 1434 | 1272 |
| EQ-Bench Creative Writing | 1829 | — |
| WildBench | — | 80.2% |
| EQ-Bench 4 | 1156 | — |
Frequently asked questions
Is GPT-5.6 Luna better than Qwen2.5 72B Instruct?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 31.9 on the Noometry Index.
Which is cheaper, GPT-5.6 Luna or Qwen2.5 72B Instruct?
GPT-5.6 Luna is cheaper. It lists at $0.20 per million input tokens and $1.20 per million output tokens; Qwen2.5 72B Instruct lists at $1.40 and $5.60.
Is GPT-5.6 Luna or Qwen2.5 72B Instruct better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 33.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 131K.
How many benchmarks do GPT-5.6 Luna and Qwen2.5 72B Instruct share?
24 benchmarks have published results for both models. GPT-5.6 Luna has 52 scored results on Noometry and Qwen2.5 72B Instruct has 43.