Model comparison
GPT-4 vs GPT-5.6 Luna
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 29.1 on the Noometry Index.
Last verified . 26 shared benchmarks.
Summary
- They share 26 benchmarks with published results for both. GPT-4 scores higher in 0 categories and GPT-5.6 Luna in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.7 to 10.8.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.1% for GPT-4 and 98.3% for GPT-5.6 Luna.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $30 / $60 for GPT-4.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 8K.
Side by side
| GPT-4 | GPT-5.6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 29.1 | 54.6 |
| Released | 2023-03-14 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 8K | 1.05M |
| Max output | 8K | 128K |
| Input $ / M tokens | $30 | $0.20 |
| Output $ / M tokens | $60 | $1.20 |
| Results tracked | 38 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
GPT-4: 31.6 (#283), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| WeirdML | 12.4% | 60.9% |
| LMArena Coding | 1254 | 1466 |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| CursorBench | — | 35.9% |
| LMArena WebDev | — | 1519 |
| SciCode | — | 53.6% |
| BigCodeBench Instruct | 46% | — |
| BigCodeBench Complete | 57.2% | — |
| ALE-Bench | — | 1,667 |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, GPT-5.6 Luna: 34.4 (#45)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | — | 43% |
| BALROG | — | 45.6% |
| GDP.pdf | — | 22.7% |
| METR Time Horizons | 36.1% | — |
| Vending-Bench 2 | — | 4,095 |
Reasoning GPT-5.6 Luna leads
GPT-4: 17.8 (#289), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| Chess Puzzles | 4% | 40% |
| LMArena Hard Prompts | 1241 | 1451 |
| Mystery Game Puzzles | 12% | 21% |
| DTBench | 62.7% | 89.1% |
| LMCA | 17.1% | 48.5% |
| Epoch Capabilities Index | 125.89 | 156.39 |
| 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% |
| Surface Evolver Bench | — | 61.9% |
| BIG-Bench Hard | 75.1% | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| WinoGrande | 87.5% | — |
Math GPT-5.6 Luna leads
GPT-4: 10.8 (#309), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 98.3% |
| LMArena Math | 1269 | 1458 |
| FrontierMath (Tiers 1-3) | — | 82.1% |
| FrontierMath Tier 4 | — | 61% |
| ProofBench | — | 60% |
| MATH Level 5 | 23% | — |
| GSM8K | 92% | — |
Knowledge GPT-5.6 Luna leads
GPT-4: 18.4 (#282), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 35.7% | 91.6% |
| LMArena Expert | 1211 | 1478 |
| SimpleQA Verified | — | 41% |
| MMLU | 86.4% | — |
| TriviaQA | 84.8% | — |
Multimodal Not comparable
GPT-4: —, GPT-5.6 Luna: 42.7 (#28)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | — | 1258 |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
| LMArena Document | — | 1457 |
Multilingual GPT-5.6 Luna leads
GPT-4: 40.6 (#215), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1246 | 1417 |
| LMArena Chinese | 1242 | 1470 |
| LMArena French | 1283 | 1456 |
| LMArena German | 1251 | 1454 |
| LMArena Japanese | 1209 | 1411 |
| LMArena Korean | 1184 | 1415 |
| LMArena Russian | 1251 | 1428 |
| LMArena Spanish | 1261 | 1448 |
Instruction Following GPT-5.6 Luna leads
GPT-4: 65.3 (#222), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1241 | 1437 |
Long Context GPT-5.6 Luna leads
GPT-4: 37.7 (#212), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1244 | 1436 |
Writing & Preference GPT-5.6 Luna leads
GPT-4: 34.9 (#268), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | GPT-4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1263 | 1431 |
| LMArena Creative Writing | 1244 | 1396 |
| EQ-Bench Creative Writing | 752 | 1829 |
| LMArena Multi-Turn | 1257 | 1434 |
| EQ-Bench 4 | — | 1156 |
Frequently asked questions
Is GPT-4 better than GPT-5.6 Luna?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 29.1 on the Noometry Index.
Which is cheaper, GPT-4 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-4 lists at $30 and $60.
Is GPT-4 or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 31.6 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 8K.
How many benchmarks do GPT-4 and GPT-5.6 Luna share?
26 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and GPT-5.6 Luna has 52.