Claude Opus 4.6 vs GPT-5
Claude Opus 4.6 
GPT-5 Verdict
- Claude Opus 4.6 leads on quality: 63.3 vs 60.8.
- Claude Opus 4.6 is stronger in agents & tools, coding, composite, human preference, knowledge, multimodal.
- GPT-5 is stronger in instruction following, long context, maths, reasoning.
- GPT-5 is 2.9× cheaper ($3.44 vs $10.00 per 1M blended).
- GPT-5 streams 2.3× faster (85 vs 37 tokens per second).
| Metric | Claude Opus 4.6 | GPT-5 |
|---|---|---|
| BenchLeader Index | 63.3 | 60.8 |
| Agents & tools score | 68.8 | 51.0 |
| Coding score | 62.3 | 60.9 |
| Composite score | 69.4 | 58.1 |
| Human preference score | 64.8 | 61.7 |
| Instruction following score | 53.9 | 65.0 |
| Knowledge score | 70.7 | 62.2 |
| Long context score | 65.5 | 65.6 |
| Maths score | 65.7 | 72.8 |
| Multimodal score | 63.6 | 59.6 |
| Reasoning score | 62.6 | 64.2 |
| Blended price $/M | $10.00 | $3.44 |
| Output speed | 37 tok/s | 85 tok/s |
| Time to first answer | 2.4 s | 60.7 s |
| Context window | 1M | 400k |
| GPQA Diamond | 90.5% | – |
| OTIS Mock AIME | 94.4% | – |
| SWE-bench Verified (Epoch) | 78.7% | – |
| Humanity's Last Exam | 19.0% | – |
| Terminal-Bench | 79.8% | 49.6% |
| SimpleBench | 67.6% | – |
| SciCode | – | 42.9% |
| Cybench | 93.0% | – |
| Remote Labor Index | 4.2% | 1.7% |
| WeirdML | 77.9% | 39.8% |
| APEX-Agents | 32.4% | 18.3% |
| FrontierCode | 26.6% | – |
| GSO-Bench | 33.3% | – |
| Epoch Capabilities Index | 155.3 | 150 |
| LMArena Text | 1497 | 1427 |
| LMArena Hard Prompts | 1527 | 1449 |
| LMArena Coding | 1546 | 1462 |
| LMArena WebDev | 1537 | – |
| LMArena Vision | 1311 | 1232 |
| AA Intelligence Index | 31.9 | 23.0 |
| IFBench | 53.1% | 73.1% |
| AA-LCR | 78.0% | 78.2% |
| MMMU-Pro | 75.4% | 74.2% |
| AA-Omniscience | 13.7 | -8.7 |
| Terminal-Bench Hard | 48.5% | 32.6% |
| GPQA Diamond (AA) | 89.6% | 85.3% |
| Humanity's Last Exam (AA) | 39.9% | 28.5% |
| τ²-Bench Telecom (AA) | 92.1% | 84.8% |
| PRBench Finance | – | 51.3% |
| PRBench Legal | – | 49.0% |
| VISTA | – | 49.7% |
| MultiNRC | – | 52.1% |
| HiL-Bench | 38.3% | – |
| TutorBench | – | 55.3% |
| EQ-Bench 4 | 1223 | – |
| Kagi LLM Benchmark | 72.4% | 72.7% |
| IFEval (HELM) | – | 87.5% |
| Omni-MATH (HELM) | – | 64.7% |
| WildBench (HELM) | – | 85.7% |
| MMLU-Pro (HELM) | – | 86.3% |
| GPQA Diamond (HELM) | – | 79.1% |
| HELM Capabilities mean | – | 80.7% |
| Aider Polyglot | – | 88.0% |
| SWE-bench Verified (bash only) | 75.6% | – |
| SWE-bench Verified (any scaffold) | 75.6% | 75.6% |
Data as of 2026-09-12. Best configuration of each model; every score links to its source on the model pages.
Claude Opus 4.6 vs GPT-5: questions
- Is Claude Opus 4.6 better than GPT-5?
- Claude Opus 4.6 leads on quality: 63.3 vs 60.8. The BenchLeader Index combines every independent quality benchmark; Claude Opus 4.6 is ahead overall as of 2026-09-12, but check the category scores for your use.
- Is Claude Opus 4.6 better than GPT-5 for coding?
- Claude Opus 4.6 scores higher in coding (62 vs 61 on the category index, where 50 is average).
- Is Claude Opus 4.6 better than GPT-5 for agentic tasks?
- Claude Opus 4.6 scores higher in agentic tasks (69 vs 51 on the category index, where 50 is average).
- Which is cheaper, Claude Opus 4.6 or GPT-5?
- GPT-5 is cheaper: $3.44 against $10.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Opus 4.6 or GPT-5?
- GPT-5 streams faster: 85 against 37 output tokens per second.
- Which has the larger context window?
- Claude Opus 4.6 accepts more context: 1M against 400k tokens.