Model comparison
GPT-5.5 vs GPT-5.6 Luna
GPT-5.5 is the stronger model overall, scoring 63.4 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 25× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.
Last verified . 50 shared benchmarks.
Summary
- They share 50 benchmarks with published results for both. GPT-5.5 scores higher in 10 categories and GPT-5.6 Luna in 0 categories; 10 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.5 leads 72.8 to 47.6.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 88.8% for GPT-5.5 and 49.1% for GPT-5.6 Luna.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $5 / $30 for GPT-5.5.
Side by side
| GPT-5.5 | GPT-5.6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 63.4 | 54.6 |
| Released | 2026-04-23 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $5 | $0.20 |
| Output $ / M tokens | $30 | $1.20 |
| Results tracked | 71 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.5 leads
GPT-5.5: 58.2 (#17), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| DeepSWE | 67% | 67.2% |
| FrontierCode | 43% | 39.8% |
| LMArena WebDev | 1513 | 1519 |
| SciCode | 56.1% | 53.6% |
| WeirdML | 84.9% | 60.9% |
| LMArena Coding | 1494 | 1466 |
| ALE-Bench | 1,943 | 1,667 |
| SWE-bench Verified | 80.6% | — |
| CursorBench | — | 35.9% |
| GSO | 40.2% | — |
| MirrorCode | 10% | — |
Agentic & Tool Use GPT-5.5 leads
GPT-5.5: 50.7 (#6), GPT-5.6 Luna: 34.4 (#45)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | 55.1% | 43% |
| GDP.pdf | 26% | 22.7% |
| Vending-Bench 2 | 7,524 | 4,095 |
| Terminal-Bench | 84.7% | — |
| OSWorld 2.0 | 13% | — |
| Remote Labor Index | 6.3% | — |
| τ²-bench Banking | 44.6% | — |
| DeepResearch Bench | 54% | — |
| PostTrainBench | 27.2% | — |
| BALROG | — | 45.6% |
| ExploitBench | 47.4% | — |
| GBAEval | 53.2% | — |
| LMArena Search | 1242 | — |
Reasoning GPT-5.5 leads
GPT-5.5: 72.8 (#11), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| ARC-AGI-2 | 85% | 59.5% |
| SimpleBench | 69% | 46.8% |
| Kagi LLM Benchmark | 88.8% | 49.1% |
| NYT Connections (extended) | 96.2% | 69.4% |
| ARC-AGI-1 | 95% | 88% |
| CritPt | 27.1% | 20.6% |
| Chess Puzzles | 54% | 40% |
| LMArena Hard Prompts | 1489 | 1451 |
| Mystery Game Puzzles | 56% | 21% |
| DTBench | 96% | 89.1% |
| LMCA | 54.3% | 48.5% |
| Surface Evolver Bench | 88.1% | 61.9% |
| Epoch Capabilities Index | 159.1 | 156.39 |
| EBR-Bench | 34.3% | — |
| Bench to the Future 3 | 0.14 | — |
| ForecastBench | 60.6 | — |
Math GPT-5.5 leads
GPT-5.5: 81.7 (#11), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| FrontierMath (Tiers 1-3) | 85.3% | 82.1% |
| FrontierMath Tier 4 | 72.5% | 61% |
| OTIS Mock AIME 2024-2025 | 100% | 98.3% |
| ProofBench | 50% | 60% |
| LMArena Math | 1486 | 1458 |
| MathArena Final-Answer Competitions | 94.3% | — |
| FrontierMath (Feb 2025 set) | 51.7% | — |
| FrontierMath Erdős | 0% | — |
| FrontierMath Tier 4 (v1) | 35.4% | — |
Knowledge GPT-5.5 leads
GPT-5.5: 64.4 (#17), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 94% | 91.6% |
| SimpleQA Verified | 63% | 41% |
| LMArena Expert | 1508 | 1478 |
| Vectara Hallucination Rate | 9.3% | — |
Multimodal GPT-5.5 leads
GPT-5.5: 46.9 (#12), GPT-5.6 Luna: 42.7 (#28)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | 1297 | 1258 |
| Blueprint-Bench 2 | 36.2% | 22.6% |
| Furniture Assembly | 44.2% | 42.5% |
| LMArena Document | 1486 | 1457 |
Multilingual GPT-5.5 leads
GPT-5.5: 56.4 (#20), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1467 | 1417 |
| LMArena Chinese | 1533 | 1470 |
| LMArena French | 1486 | 1456 |
| LMArena German | 1480 | 1454 |
| LMArena Japanese | 1498 | 1411 |
| LMArena Korean | 1460 | 1415 |
| LMArena Russian | 1473 | 1428 |
| LMArena Spanish | 1468 | 1448 |
Instruction Following GPT-5.5 leads
GPT-5.5: 77.5 (#18), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1479 | 1437 |
Long Context GPT-5.5 leads
GPT-5.5: 48.3 (#12), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1484 | 1436 |
| CL-bench Life | 22.2% | — |
Writing & Preference GPT-5.5 leads
GPT-5.5: 72.7 (#13), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | GPT-5.5 | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1472 | 1431 |
| LMArena Creative Writing | 1455 | 1396 |
| EQ-Bench Creative Writing | 1844 | 1829 |
| EQ-Bench 4 | 1315 | 1156 |
| LMArena Multi-Turn | 1476 | 1434 |
Frequently asked questions
Is GPT-5.5 better than GPT-5.6 Luna?
GPT-5.5 is the stronger model overall, scoring 63.4 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 25× less per token, which makes it the better buy when GPT-5.5's lead doesn't matter for your workload.
Which is cheaper, GPT-5.5 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-5.5 lists at $5 and $30.
Is GPT-5.5 or GPT-5.6 Luna better for coding?
GPT-5.5 scores higher on coding benchmarks: 58.2 versus 54.5 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do GPT-5.5 and GPT-5.6 Luna share?
50 benchmarks have published results for both models. GPT-5.5 has 71 scored results on Noometry and GPT-5.6 Luna has 52.