Claude Opus 4.6 vs GPT-5.4 Pro
Claude Opus 4.6 
GPT-5.4 Pro Verdict
- GPT-5.4 Pro leads on quality: 64.2 vs 63.1.
- Claude Opus 4.6 is stronger in agents & tools, coding, composite, human preference, long context, maths.
- GPT-5.4 Pro is stronger in instruction following, knowledge, multimodal, reasoning.
- Claude Opus 4.6 is 6.8× cheaper ($10.00 vs $67.50 per 1M blended).
- Claude Opus 4.6 streams 37.9× faster (38 vs 1 tokens per second).
| Metric | Claude Opus 4.6 | GPT-5.4 Pro |
|---|---|---|
| BenchLeader Index | 63.1 | 64.2 |
| Agents & tools score | 68.8 | – |
| Coding score | 62.3 | – |
| Composite score | 69.2 | – |
| Human preference score | 64.6 | – |
| Instruction following score | 53.7 | 67.3 |
| Knowledge score | 70.5 | 73.2 |
| Long context score | 65.5 | – |
| Maths score | 65.6 | – |
| Multimodal score | 63.6 | 69.8 |
| Reasoning score | 62.3 | 75.1 |
| Blended price $/M | $10.00 | $67.50 |
| Output speed | 38 tok/s | 1 tok/s |
| Time to first answer | 2.0 s | 6.7 s |
| Context window | 1M | 1.1M |
| GPQA Diamond | 90.5% | – |
| OTIS Mock AIME | 94.4% | – |
| SWE-bench Verified (Epoch) | 78.7% | – |
| Humanity's Last Exam | 19.0% | 44.3% |
| Terminal-Bench | 79.8% | – |
| SimpleBench | 67.6% | 74.1% |
| Cybench | 93.0% | – |
| Remote Labor Index | 4.2% | – |
| WeirdML | 77.9% | – |
| APEX-Agents | 32.4% | – |
| FrontierCode | 26.6% | – |
| GSO-Bench | 33.3% | – |
| Epoch Capabilities Index | 155.3 | 158.9 |
| LMArena Text | 1498 | – |
| LMArena Hard Prompts | 1527 | – |
| LMArena Coding | 1546 | – |
| LMArena WebDev | 1537 | – |
| LMArena Vision | 1311 | – |
| AA Intelligence Index | 31.9 | – |
| IFBench | 53.1% | – |
| AA-LCR | 78.0% | – |
| MMMU-Pro | 75.4% | – |
| AA-Omniscience | 13.7 | – |
| Terminal-Bench Hard | 48.5% | – |
| GPQA Diamond (AA) | 89.6% | – |
| Humanity's Last Exam (AA) | 39.9% | – |
| τ²-Bench Telecom (AA) | 92.1% | – |
| MultiChallenge | – | 69.2% |
| VISTA | – | 53.9% |
| MultiNRC | – | 62.3% |
| HiL-Bench | 38.3% | – |
| TutorBench | – | 56.6% |
| EQ-Bench 4 | 1223 | – |
| Kagi LLM Benchmark | 72.4% | – |
| SWE-bench Verified (bash only) | 75.6% | – |
| SWE-bench Verified (any scaffold) | 75.6% | – |
Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.