Model comparison
Claude Opus 4 vs GPT-5.6 Luna
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 43.1 on the Noometry Index.
Last verified . 30 shared benchmarks.
Summary
- They share 30 benchmarks with published results for both. Claude Opus 4 scores higher in 2 categories and GPT-5.6 Luna in 8 categories; 9 gaps are clear of the uncertainty.
- The widest gap is in math, where GPT-5.6 Luna leads 77.7 to 42.0.
- The biggest single-benchmark swing is ARC-AGI-1: 35.7% for Claude Opus 4 and 88% for GPT-5.6 Luna.
- GPT-5.6 Luna is cheaper at $0.20 / $1.20 per million input/output tokens, against $15 / $75 for Claude Opus 4.
- GPT-5.6 Luna accepts more context: 1.05M tokens versus 200K.
Side by side
| Claude Opus 4 | GPT-5.6 Luna | |
|---|---|---|
| Provider | Anthropic | OpenAI |
| Noometry Index | 43.1 | 54.6 |
| Released | 2025-05-22 | 2026-07-09 |
| Weights | Proprietary | Proprietary |
| Context window | 200K | 1.05M |
| Max output | 32K | 128K |
| Input $ / M tokens | $15 | $0.20 |
| Output $ / M tokens | $75 | $1.20 |
| Results tracked | 56 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.6 Luna leads
Claude Opus 4: 47.2 (#62), GPT-5.6 Luna: 54.5 (#28)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| WeirdML | 43.7% | 60.9% |
| LMArena Coding | 1442 | 1466 |
| SWE-bench Verified | 70.7% | — |
| DeepSWE | — | 67.2% |
| FrontierCode | — | 39.8% |
| SWE-bench Verified (bash only) | 67.6% | — |
| Aider Polyglot | 72% | — |
| CursorBench | — | 35.9% |
| LMArena WebDev | — | 1519 |
| SciCode | — | 53.6% |
| GSO | 6.9% | — |
| ALE-Bench | — | 1,667 |
| AlgoTune | 1.33 | — |
Agentic & Tool Use Too close to call
Claude Opus 4: 34.8 (#42), GPT-5.6 Luna: 34.4 (#45)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| APEX-Agents | — | 43% |
| Cybench | 38% | — |
| DeepResearch Bench | 46.8% | — |
| BALROG | — | 45.6% |
| GDP.pdf | — | 22.7% |
| LMArena Search | 1127 | — |
| METR Time Horizons | 63.9% | — |
| Vending-Bench 2 | — | 4,095 |
Reasoning GPT-5.6 Luna leads
Claude Opus 4: 27.3 (#121), GPT-5.6 Luna: 47.6 (#43)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| ARC-AGI-2 | 8.6% | 59.5% |
| SimpleBench | 58.8% | 46.8% |
| Kagi LLM Benchmark | 74.3% | 49.1% |
| ARC-AGI-1 | 35.7% | 88% |
| CritPt | 0.3% | 20.6% |
| LMArena Hard Prompts | 1399 | 1451 |
| DTBench | 81.6% | 89.1% |
| LMCA | 37.4% | 48.5% |
| Epoch Capabilities Index | 142.67 | 156.39 |
| NYT Connections (extended) | — | 69.4% |
| Chess Puzzles | — | 40% |
| EnigmaEval | 5.6% | — |
| Mystery Game Puzzles | — | 21% |
| Surface Evolver Bench | — | 61.9% |
| ForecastBench | 61.1 | — |
Math GPT-5.6 Luna leads
Claude Opus 4: 42.0 (#86), GPT-5.6 Luna: 77.7 (#14)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 64.4% | 98.3% |
| LMArena Math | 1390 | 1458 |
| FrontierMath (Tiers 1-3) | — | 82.1% |
| FrontierMath Tier 4 | — | 61% |
| ProofBench | — | 60% |
| Omni-MATH | 61.6% | — |
| MATH Level 5 | 85% | — |
| FrontierMath (Feb 2025 set) | 4.5% | — |
| FrontierMath Tier 4 (v1) | 4.2% | — |
Knowledge GPT-5.6 Luna leads
Claude Opus 4: 44.0 (#88), GPT-5.6 Luna: 58.5 (#34)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| GPQA Diamond | 76.3% | 91.6% |
| LMArena Expert | 1386 | 1478 |
| Humanity's Last Exam | 10.7% | — |
| SimpleQA Verified | — | 41% |
| MMLU-Pro | 87.5% | — |
| Confabulations | 15.9% | — |
| Vectara Hallucination Rate | 12% | — |
| GPQA (HELM) | 70.8% | — |
Multimodal GPT-5.6 Luna leads
Claude Opus 4: 31.5 (#106), GPT-5.6 Luna: 42.7 (#28)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Vision | 1192 | 1258 |
| GeoBench | 49% | — |
| VPCT | 38% | — |
| Blueprint-Bench 2 | — | 22.6% |
| Furniture Assembly | — | 42.5% |
| LMArena Document | — | 1457 |
Multilingual GPT-5.6 Luna leads
Claude Opus 4: 48.8 (#138), GPT-5.6 Luna: 52.8 (#78)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Non-English | 1362 | 1417 |
| LMArena Chinese | 1386 | 1470 |
| LMArena French | 1372 | 1456 |
| LMArena German | 1391 | 1454 |
| LMArena Japanese | 1331 | 1411 |
| LMArena Korean | 1321 | 1415 |
| LMArena Russian | 1392 | 1428 |
| LMArena Spanish | 1389 | 1448 |
Instruction Following Claude Opus 4 leads
Claude Opus 4: 77.1 (#28), GPT-5.6 Luna: 75.6 (#57)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Instruction Following | 1406 | 1437 |
| IFEval | 91.8% | — |
Long Context GPT-5.6 Luna leads
Claude Opus 4: 39.6 (#172), GPT-5.6 Luna: 43.9 (#82)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Longer Query | 1422 | 1436 |
| Fiction.LiveBench | 61.1% | — |
Writing & Preference GPT-5.6 Luna leads
Claude Opus 4: 61.2 (#89), GPT-5.6 Luna: 68.0 (#29)
| Benchmark | Claude Opus 4 | GPT-5.6 Luna |
|---|---|---|
| LMArena Text | 1377 | 1431 |
| LMArena Creative Writing | 1387 | 1396 |
| EQ-Bench Creative Writing | 1580 | 1829 |
| LMArena Multi-Turn | 1396 | 1434 |
| Short-Story Creative Writing | 83.6% | — |
| WildBench | 85.2% | — |
| EQ-Bench 4 | — | 1156 |
Frequently asked questions
Is Claude Opus 4 better than GPT-5.6 Luna?
GPT-5.6 Luna is the stronger model overall, scoring 54.6 to 43.1 on the Noometry Index.
Which is cheaper, Claude Opus 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; Claude Opus 4 lists at $15 and $75.
Is Claude Opus 4 or GPT-5.6 Luna better for coding?
GPT-5.6 Luna scores higher on coding benchmarks: 54.5 versus 47.2 in the Noometry coding category.
Which has the bigger context window?
GPT-5.6 Luna does, with 1.05M tokens against 200K.
How many benchmarks do Claude Opus 4 and GPT-5.6 Luna share?
30 benchmarks have published results for both models. Claude Opus 4 has 56 scored results on Noometry and GPT-5.6 Luna has 52.