Model comparison
GPT-5.6 Luna vs GPT-6 Sol
GPT-6 Sol is the stronger model overall, scoring 61.8 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 8.9× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.
Last verified . 43 shared benchmarks.
Summary
- They share 43 benchmarks with published results for both. GPT-5.6 Luna scores higher in 3 categories and GPT-6 Sol in 7 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where GPT-6 Sol leads 74.0 to 47.6.
- The biggest single-benchmark swing is Mystery Game Puzzles: 21% for GPT-5.6 Luna and 56% for GPT-6 Sol.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $2 / $10 for GPT-6 Sol.
Side by side
| GPT-5.6 Luna | GPT-6 Sol | |
|---|---|---|
| Provider | OpenAI | OpenAI |
| Noometry Index | 54.6 | 61.8 |
| Released | 2026-07-09 | 2026-09-22 |
| Weights | Proprietary | Proprietary |
| Context window | 1.05M | 1.05M |
| Max output | 128K | 128K |
| Input $ / M tokens | $0.20 | $2 |
| Output $ / M tokens | $1.20 | $10 |
| Results tracked | 52 | 45 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-6 Sol leads
GPT-5.6 Luna: 54.5 (#28), GPT-6 Sol: 60.1 (#11)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| DeepSWE | 67.2% | 68.8% |
| FrontierCode | 39.8% | 49.3% |
| LMArena WebDev | 1519 | 1688 |
| SciCode | 53.6% | 57.6% |
| LMArena Coding | 1466 | 1447 |
| ALE-Bench | 1,667 | 2,462 |
| CursorBench | 35.9% | — |
| WeirdML | 60.9% | — |
Agentic & Tool Use GPT-6 Sol leads
GPT-5.6 Luna: 34.4 (#45), GPT-6 Sol: 37.2 (#36)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| APEX-Agents | 43% | 54.3% |
| GDP.pdf | 22.7% | 26.4% |
| Vending-Bench 2 | 4,095 | 14,428 |
| BALROG | 45.6% | — |
Reasoning GPT-6 Sol leads
GPT-5.6 Luna: 47.6 (#43), GPT-6 Sol: 74.0 (#9)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| ARC-AGI-2 | 59.5% | 89.6% |
| NYT Connections (extended) | 69.4% | 90.1% |
| ARC-AGI-1 | 88% | 95.5% |
| CritPt | 20.6% | 30.9% |
| LMArena Hard Prompts | 1451 | 1418 |
| Mystery Game Puzzles | 21% | 56% |
| DTBench | 89.1% | 97.3% |
| LMCA | 48.5% | 59.1% |
| Epoch Capabilities Index | 156.39 | 162.72 |
| SimpleBench | 46.8% | — |
| Kagi LLM Benchmark | 49.1% | — |
| Chess Puzzles | 40% | — |
| EBR-Bench | — | 53.3% |
| Surface Evolver Bench | 61.9% | — |
Math GPT-6 Sol leads
GPT-5.6 Luna: 77.7 (#14), GPT-6 Sol: 87.2 (#7)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| FrontierMath (Tiers 1-3) | 82.1% | 89.8% |
| FrontierMath Tier 4 | 61% | 90% |
| OTIS Mock AIME 2024-2025 | 98.3% | 100% |
| ProofBench | 60% | 83% |
| LMArena Math | 1458 | 1402 |
Knowledge GPT-6 Sol leads
GPT-5.6 Luna: 58.5 (#34), GPT-6 Sol: 64.8 (#15)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| GPQA Diamond | 91.6% | 94.3% |
| SimpleQA Verified | 41% | 60.7% |
| LMArena Expert | 1478 | 1439 |
| Vectara Hallucination Rate | — | 6.5% |
Multimodal GPT-6 Sol leads
GPT-5.6 Luna: 42.7 (#28), GPT-6 Sol: 47.6 (#10)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| LMArena Vision | 1258 | 1245 |
| Blueprint-Bench 2 | 22.6% | 36.9% |
| Furniture Assembly | 42.5% | 58.3% |
| LMArena Document | 1457 | — |
Multilingual GPT-5.6 Luna leads
GPT-5.6 Luna: 52.8 (#78), GPT-6 Sol: 50.5 (#118)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| LMArena Non-English | 1417 | 1385 |
| LMArena Chinese | 1470 | 1405 |
| LMArena French | 1456 | 1410 |
| LMArena German | 1454 | 1390 |
| LMArena Japanese | 1411 | 1385 |
| LMArena Korean | 1415 | 1341 |
| LMArena Russian | 1428 | 1401 |
| LMArena Spanish | 1448 | 1384 |
Instruction Following GPT-5.6 Luna leads
GPT-5.6 Luna: 75.6 (#57), GPT-6 Sol: 74.5 (#94)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| LMArena Instruction Following | 1437 | 1412 |
Long Context Too close to call
GPT-5.6 Luna: 43.9 (#82), GPT-6 Sol: 43.1 (#108)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| LMArena Longer Query | 1436 | 1411 |
Writing & Preference GPT-6 Sol leads
GPT-5.6 Luna: 68.0 (#29), GPT-6 Sol: 71.9 (#18)
| Benchmark | GPT-5.6 Luna | GPT-6 Sol |
|---|---|---|
| LMArena Text | 1431 | 1395 |
| LMArena Creative Writing | 1396 | 1378 |
| EQ-Bench Creative Writing | 1829 | 2125 |
| LMArena Multi-Turn | 1434 | 1412 |
| EQ-Bench 4 | 1156 | — |
Frequently asked questions
Is GPT-5.6 Luna better than GPT-6 Sol?
GPT-6 Sol is the stronger model overall, scoring 61.8 to 54.6 on the Noometry Index. GPT-5.6 Luna costs 8.9× less per token, which makes it the better buy when GPT-6 Sol's lead doesn't matter for your workload.
Which is cheaper, GPT-5.6 Luna or GPT-6 Sol?
GPT-5.6 Luna is cheaper. It lists at $0.20 per million input tokens and $1.20 per million output tokens; GPT-6 Sol lists at $2 and $10.
Is GPT-5.6 Luna or GPT-6 Sol better for coding?
GPT-6 Sol scores higher on coding benchmarks: 60.1 versus 54.5 in the Noometry coding category.
Which has the bigger context window?
Both accept 1.05M tokens.
How many benchmarks do GPT-5.6 Luna and GPT-6 Sol share?
43 benchmarks have published results for both models. GPT-5.6 Luna has 52 scored results on Noometry and GPT-6 Sol has 45.