Claude Opus 5 vs GPT-6 Luna
Claude Opus 5 
GPT-6 Luna Verdict
- Claude Opus 5 (high) leads on quality: 69.5 vs 61.2.
- Claude Opus 5 (high) is stronger in agents & tools, coding, composite, human preference, knowledge, maths, multimodal.
- GPT-6 Luna (max) is stronger in long context, reasoning.
- GPT-6 Luna (max) is 50× cheaper ($0.200 vs $10.00 per 1M blended).
- GPT-6 Luna (max) streams 2.9× faster (157 vs 55 tokens per second).
| Metric | Claude Opus 5 (high) | GPT-6 Luna (max) |
|---|---|---|
| BenchLeader Index | 69.5 | 61.2 |
| Agents & tools score | 66.0 | – |
| Coding score | 71.4 | – |
| Composite score | 86.9 | 73.9 |
| Human preference score | 69.8 | – |
| Knowledge score | 78.5 | 63.2 |
| Long context score | 65.3 | 67.5 |
| Maths score | 72.3 | – |
| Multimodal score | 67.2 | 59.2 |
| Reasoning score | 74.7 | 76.1 |
| Blended price $/M | $10.00 | $0.200 |
| Output speed | 55 tok/s | 157 tok/s |
| Time to first answer | 19.8 s | 142.6 s |
| Context window | 1M | 1.1M |
| Terminal-Bench | 50.3% | – |
| OSWorld-Verified 2.0 | 29.0% | – |
| SciCode | 54.3% | – |
| WeirdML | 91.6% | – |
| LMArena Text | 1493 | – |
| LMArena Hard Prompts | 1517 | – |
| LMArena Coding | 1533 | – |
| LMArena WebDev | 1660 | – |
| LMArena Vision | 1321 | – |
| LMArena Agent | 10.2 | – |
| AA Intelligence Index | 48.1 | 37.3 |
| AA-LCR | 79.0% | 83.3% |
| MMMU-Pro | 82.4% | 75.5% |
| AA-Omniscience | 33.7 | 0.7 |
| GPQA Diamond (AA) | 93.7% | – |
| Humanity's Last Exam (AA) | 52.8% | 38.5% |
| SciCode (AA) | 55.4% | 54.6% |
| ARC-AGI-1 | 97.5% | – |
| ARC-AGI-2 | 88.3% | – |
| ARC-AGI-3 | 30.2% | – |
| CritPt | 28.3% | 19.4% |
| GDPval (AA) | 54.0% | 43.4% |
| τ²-Bench Banking (AA) | 44.7% | – |
| LMArena Maths | 1525 | – |
| LMArena Creative Writing | 1474 | – |
| LMArena Instruction Following | 1499 | – |
| LMArena Multi-turn | 1484 | – |
| LMArena Longer Queries | 1508 | – |
| LMArena Document | 1490 | – |
| DeepSWE | 72.8% | – |
| LMCA | 63.7% | – |
| DTBench | 97.9% | – |
| CursorBench | 44.7% | – |
| ALE-Bench | 2164.6 | – |
| Terminal-Bench 4.0 (AA) | 46.0% | 12.6% |
| Terminal-Bench 2.1 (AA) | 87.6% | – |
| AutomationBench | 53.6% | 53.2% |
| GDP.pdf | 19.6% | 20.4% |
| MLCR | 59.4% | 16.1% |
| AA-Omniscience: accuracy | 58.9% | 43.8% |
| AA-Omniscience: non-hallucination | 38.8% | 23.3% |
| AA-Briefcase | 1573 | 1299 |
Data as of 2026-09-23. Best configuration of each model; every score links to its source on the model pages.
Claude Opus 5 vs GPT-6 Luna: questions
- Is Claude Opus 5 better than GPT-6 Luna?
- Claude Opus 5 (high) leads on quality: 69.5 vs 61.2. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5 (high) is ahead overall as of 2026-09-23, but check the category scores for your use.
- Which is cheaper, Claude Opus 5 or GPT-6 Luna?
- GPT-6 Luna is cheaper: $0.200 against $10.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Opus 5 or GPT-6 Luna?
- GPT-6 Luna streams faster: 157 against 55 output tokens per second.
- Which has the larger context window?
- GPT-6 Luna accepts more context: 1.1M against 1M tokens.