Model comparison
GPT-5.4 Pro vs GPT-5.6 Luna
GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 150× less per token, which makes it the better buy when GPT-5.4 Pro's lead doesn't matter for your workload.
Last verified . 11 shared benchmarks.
Summary
- They share 11 benchmarks with published results for both. GPT-5.4 Pro scores higher in 2 categories and GPT-5.6 Luna in 2 categories; 4 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-5.4 Pro leads 70.7 to 47.6.
- The biggest single-benchmark swing is SimpleBench: 74.1% for GPT-5.4 Pro and 46.8% for GPT-5.6 Luna.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $30 / $180 for GPT-5.4 Pro.
Side by side
| GPT-5.4 Pro | GPT-5.6 Luna | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 58.9 | 54.6 |
| Released | 2026-03-05 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $30 | $0.20 |
| Output $ / M tokens | $180 | $1.20 |
| Results tracked | 16 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
GPT-5.4 Pro: 43.2 (#90), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| WeirdML | 57.4% | 60.9% |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| CursorBench | — | 35.9% |
| LMArena WebDev | — | 1519 |
| SciCode | — | 53.6% |
| LMArena Coding | — | 1466 |
| ALE-Bench | — | 1,667 |
Agentic & Tool Use Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 34.4 (#45)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | — | 43% |
| BALROG | — | 45.6% |
| GDP.pdf | — | 22.7% |
| Vending-Bench 2 | — | 4,095 |
Reasoning GPT-5.4 Pro leads
GPT-5.4 Pro: 70.7 (#13), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| ARC-AGI-2 | 83.3% | 59.5% |
| SimpleBench | 74.1% | 46.8% |
| ARC-AGI-1 | 94.5% | 88% |
| CritPt | 30% | 20.6% |
| Chess Puzzles | 58.6% | 40% |
| Epoch Capabilities Index | 158.93 | 156.39 |
| Kagi LLM Benchmark | — | 49.1% |
| NYT Connections (extended) | — | 69.4% |
| EnigmaEval | 23.8% | — |
| LMArena Hard Prompts | — | 1451 |
| Mystery Game Puzzles | — | 21% |
| DTBench | — | 89.1% |
| LMCA | — | 48.5% |
| Surface Evolver Bench | — | 61.9% |
Math GPT-5.6 Luna leads
GPT-5.4 Pro: 72.4 (#22), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| FrontierMath (Tiers 1-3) | 82.5% | 82.1% |
| FrontierMath Tier 4 | 58.5% | 61% |
| OTIS Mock AIME 2024-2025 | — | 98.3% |
| ProofBench | — | 60% |
| LMArena Math | — | 1458 |
| FrontierMath (Feb 2025 set) | 50% | — |
| FrontierMath Tier 4 (v1) | 37.5% | — |
Knowledge GPT-5.4 Pro leads
GPT-5.4 Pro: 68.3 (#7), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 94.6% | 91.6% |
| SimpleQA Verified | 46.3% | 41% |
| Humanity's Last Exam | 44.3% | — |
| Vectara Hallucination Rate | 8.3% | — |
| LMArena Expert | — | 1478 |
Multimodal Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 42.7 (#28)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | — | 1258 |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
| LMArena Document | — | 1457 |
Multilingual Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 52.8 (#78)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | — | 1417 |
| LMArena Chinese | — | 1470 |
| LMArena French | — | 1456 |
| LMArena German | — | 1454 |
| LMArena Japanese | — | 1411 |
| LMArena Korean | — | 1415 |
| LMArena Russian | — | 1428 |
| LMArena Spanish | — | 1448 |
Instruction Following Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 75.6 (#57)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | — | 1437 |
Long Context Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 43.9 (#82)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | — | 1436 |
Writing & Preference Not comparable
GPT-5.4 Pro: —, GPT-5.6 Luna: 68.0 (#29)
| Benchmark | GPT-5.4 Pro | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | — | 1431 |
| LMArena Creative Writing | — | 1396 |
| EQ-Bench Creative Writing | — | 1829 |
| EQ-Bench 4 | — | 1156 |
| LMArena Multi-Turn | — | 1434 |
Frequently asked questions
Is GPT-5.4 Pro better than GPT-5.6 Luna?
GPT-5.4 Pro is the stronger model overall, scoring 58.9 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 150× less per token, which makes it the better buy when GPT-5.4 Pro's lead doesn't matter for your workload.
Which is cheaper, GPT-5.4 Pro 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-5.4 Pro lists at $30 and $180.
Is GPT-5.4 Pro or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 43.2 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do GPT-5.4 Pro and GPT-5.6 Luna share?
11 benchmarks have published results for both models. GPT-5.4 Pro has 16 scored results on Noometry and GPT-5.6 Luna has 52.