Model comparison
GPT-5 vs GPT-5.6 Luna
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 50.9 on the Noometry Index.
Last verified . 40 shared benchmarks.
Summary
- They share 40 benchmarks with published results for both. GPT-5 scores higher in 2 categories and GPT-5.6 Luna in 8 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in long context, where GPT-5 leads 69.5 to 43.9.
- The biggest single-benchmark swing is ARC-AGI-2: 9.9% for GPT-5 and 59.5% for GPT-5.6 Luna.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $1.25 / $10 for GPT-5.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 400K.
Side by side
| GPT-5 | GPT-5.6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 50.9 | 54.6 |
| Released | 2025-08-07 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 400K | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $1.25 | $0.20 |
| Output $ / M tokens | $10 | $1.20 |
| Results tracked | 69 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
GPT-5: 50.3 (#47), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena WebDev | 1418 | 1519 |
| SciCode | 42.9% | 53.6% |
| WeirdML | 60.7% | 60.9% |
| LMArena Coding | 1436 | 1466 |
| ALE-Bench | 1,162 | 1,667 |
| SWE-bench Verified | 73.6% | — |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| SWE-bench Verified (bash only) | 65% | — |
| Aider Polyglot | 88% | — |
| CursorBench | — | 35.9% |
| GSO | 6.9% | — |
| AlgoTune | 1.67 | — |
Agentic & Tool Use GPT-5.6 Luna leads
GPT-5: 33.1 (#56), GPT-5.6 Luna: 34.4 (#45)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| BALROG | 32.8% | 45.6% |
| Terminal-Bench | 49.6% | — |
| APEX-Agents | — | 43% |
| GDPval | 34.8% | — |
| Remote Labor Index | 1.7% | — |
| DeepResearch Bench | 49.6% | — |
| GDP.pdf | — | 22.7% |
| LMArena Search | 1133 | — |
| METR Time Horizons | 69.6% | — |
| Vending-Bench 2 | — | 4,095 |
Reasoning GPT-5.6 Luna leads
GPT-5: 38.3 (#64), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| ARC-AGI-2 | 9.9% | 59.5% |
| SimpleBench | 56.7% | 46.8% |
| Kagi LLM Benchmark | 72.7% | 49.1% |
| ARC-AGI-1 | 65.7% | 88% |
| CritPt | 12.6% | 20.6% |
| Chess Puzzles | 37% | 40% |
| LMArena Hard Prompts | 1416 | 1451 |
| Mystery Game Puzzles | 23% | 21% |
| DTBench | 90.7% | 89.1% |
| LMCA | 40% | 48.5% |
| Epoch Capabilities Index | 150 | 156.39 |
| NYT Connections (extended) | — | 69.4% |
| EnigmaEval | 10.5% | — |
| EBR-Bench | 12.7% | — |
| Surface Evolver Bench | — | 61.9% |
| ForecastBench | 61.4 | — |
Math GPT-5.6 Luna leads
GPT-5: 55.0 (#44), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| FrontierMath (Tiers 1-3) | 55.4% | 82.1% |
| FrontierMath Tier 4 | 22% | 61% |
| OTIS Mock AIME 2024-2025 | 91.4% | 98.3% |
| ProofBench | 18% | 60% |
| LMArena Math | 1407 | 1458 |
| Omni-MATH | 64.7% | — |
| MATH Level 5 | 98.1% | — |
| FrontierMath (Feb 2025 set) | 32.4% | — |
| FrontierMath Tier 4 (v1) | 12.5% | — |
Knowledge GPT-5.6 Luna leads
GPT-5: 56.6 (#43), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 86.2% | 91.6% |
| SimpleQA Verified | 50.1% | 41% |
| LMArena Expert | 1419 | 1478 |
| Humanity's Last Exam | 25.3% | — |
| MMLU-Pro | 86.3% | — |
| Confabulations | 10.3% | — |
| Vectara Hallucination Rate | 14.7% | — |
| GPQA (HELM) | 79.2% | — |
Multimodal GPT-5 leads
GPT-5: 46.8 (#13), GPT-5.6 Luna: 42.7 (#28)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | 1232 | 1258 |
| GeoBench | 81% | — |
| VPCT | 66% | — |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
| LMArena Document | — | 1457 |
Multilingual GPT-5.6 Luna leads
GPT-5: 51.4 (#110), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1397 | 1417 |
| LMArena Chinese | 1422 | 1470 |
| LMArena French | 1410 | 1456 |
| LMArena German | 1416 | 1454 |
| LMArena Japanese | 1409 | 1411 |
| LMArena Korean | 1360 | 1415 |
| LMArena Russian | 1406 | 1428 |
| LMArena Spanish | 1399 | 1448 |
Instruction Following GPT-5.6 Luna leads
GPT-5: 73.8 (#113), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1388 | 1437 |
| IFEval | 87.5% | — |
Long Context GPT-5 leads
GPT-5: 69.5 (#2), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1399 | 1436 |
| Fiction.LiveBench | 97.2% | — |
Writing & Preference GPT-5.6 Luna leads
GPT-5: 63.4 (#65), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | GPT-5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1406 | 1431 |
| LMArena Creative Writing | 1365 | 1396 |
| EQ-Bench Creative Writing | 1627 | 1829 |
| LMArena Multi-Turn | 1426 | 1434 |
| Short-Story Creative Writing | 86% | — |
| WildBench | 85.7% | — |
| EQ-Bench 4 | — | 1156 |
Frequently asked questions
Is GPT-5 better than GPT-5.6 Luna?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 50.9 on the Noometry Index.
Which is cheaper, GPT-5 or GPT-5.6 Luna?
GPT-5.6 Luna is cheaper. It lists at $0.20 per million input tokens and $1.20 per million output tokens; GPT-5 lists at $1.25 and $10.
Is GPT-5 or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 50.3 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 400K.
How many benchmarks do GPT-5 and GPT-5.6 Luna share?
40 benchmarks have published results for both models. GPT-5 has 69 scored results on Noometry and GPT-5.6 Luna has 52.