Claude Opus 5.5 vs GPT-5.4
Claude Opus 5.5 
GPT-5.4 Verdict
- Claude Opus 5.5 (thinking) leads on quality: 70.7 vs 65.1.
- Claude Opus 5.5 (thinking) is stronger in composite, knowledge, long context, multimodal, reasoning.
- GPT-5.4 (xhigh) is stronger in agents & tools, coding, instruction following, maths.
- GPT-5.4 (xhigh) is 1.4× cheaper ($5.63 vs $8.00 per 1M blended).
- GPT-5.4 (xhigh) streams 2.9× faster (140 vs 48 tokens per second).
| Metric | Claude Opus 5.5 (thinking) | GPT-5.4 (xhigh) |
|---|---|---|
| BenchLeader Index | 70.7 | 65.1 |
| Composite score | 95.0 | 68.4 |
| Knowledge score | 85.0 | 60.4 |
| Long context score | 68.5 | 67.2 |
| Multimodal score | 71.9 | 62.4 |
| Reasoning score | 95.0 | 71.2 |
| Agents & tools score | – | 63.5 |
| Coding score | – | 66.4 |
| Instruction following score | – | 72.1 |
| Maths score | – | 66.4 |
| Blended price $/M | $8.00 | $5.63 |
| Output speed | 48 tok/s | 140 tok/s |
| Time to first answer | 4.2 s | 149.5 s |
| Context window | 1M | 1.1M |
| GPQA Diamond | – | 93.3% |
| FrontierMath Tiers 1–3 | – | 78.6% |
| FrontierMath Tier 4 | – | 49.0% |
| OTIS Mock AIME | – | 95.3% |
| SimpleQA Verified | – | 45.1% |
| Humanity's Last Exam | – | 36.2% |
| SciCode | – | 56.6% |
| WeirdML | – | 77.7% |
| ProofBench | – | 56.0% |
| GSO-Bench | – | 31.4% |
| LiveBench | – | 78.0% |
| LiveBench Reasoning | – | 88.1% |
| LiveBench Coding | – | 77.5% |
| LiveBench Agentic Coding | – | 53.8% |
| LiveBench Mathematics | – | 94.2% |
| LiveBench Data Analysis | – | 79.3% |
| LiveBench Language | – | 82.6% |
| LiveBench Instruction Following | – | 70.2% |
| AA Intelligence Index | 57.6 | 39.0 |
| IFBench | – | 74.0% |
| AA-LCR | 84.7% | 82.0% |
| MMMU-Pro | 87.7% | 78.4% |
| AA-Omniscience | 46.4 | 5.8 |
| Terminal-Bench Hard | – | 57.6% |
| GPQA Diamond (AA) | – | 92.0% |
| Humanity's Last Exam (AA) | 61.4% | 43.7% |
| SciCode (AA) | 66.9% | – |
| τ²-Bench Telecom (AA) | – | 87.1% |
| AIME (Vals) | – | 96.7% |
| LiveCodeBench | – | 84.1% |
| MMLU-Pro | – | 87.5% |
| LegalBench | – | 86.0% |
| CorpFin | – | 65.3% |
| TaxEval | – | 74.0% |
| MedQA | – | 96.1% |
| SWE-bench (Vals) | – | 78.2% |
| GPQA Diamond (Vals) | – | 91.7% |
| AIME 2026 | – | 99.2% |
| HMMT February 2026 | – | 97.7% |
| MathArena Apex | – | 54.2% |
| SWE-Bench Pro | – | 59.1% |
| MCP Atlas | – | 70.6% |
| ARC-AGI-1 | – | 93.7% |
| ARC-AGI-2 | – | 74.0% |
| CritPt | 31.7% | 23.4% |
| GDPval (AA) | 67.3% | 36.6% |
| τ²-Bench Banking (AA) | – | 39.6% |
| APEX-Agents (AA) | – | 33.3% |
| CaseLaw v2 | – | 63.8% |
| Code Migration | – | 35.0% |
| Harvey's Legal Agent Benchmark | – | 0.0% |
| MedCode | – | 41.3% |
| MedScribe | – | 77.5% |
| MMMU-Pro (Vals) | – | 87.5% |
| MortgageTax | – | 68.3% |
| ProgramBench | – | 0.0% |
| SAGE | – | 43.3% |
| SkillsBench | – | 51.7% |
| Vibe Code Bench v1.1 | – | 67.4% |
| SWE-Bench Pro (private) | – | 43.4% |
| SWE Atlas: Codebase QnA | – | 40.8% |
| SWE Atlas: Refactoring | – | 44.3% |
| SWE Atlas: Test Writing | – | 44.4% |
| Chess Puzzles | – | 44.0% |
| EBR-bench | – | 25.4% |
| Mystery Game Puzzles | – | 37.0% |
| CL-bench | – | 27.9% |
| CL-bench Life | – | 21.7% |
| METR Time Horizons | – | 74.3% |
| DeepSWE | – | 51.8% |
| LMCA | – | 52.0% |
| DTBench | – | 94.4% |
| Terminal-Bench 4.0 (AA) | 59.6% | – |
| Terminal-Bench 2.1 (AA) | – | 78.3% |
| AutomationBench | 69.5% | – |
| GDP.pdf | 26.2% | – |
| Harvey LAB | 91.2% | – |
| AA-Omniscience: accuracy | 66.2% | 50.9% |
| AA-Omniscience: non-hallucination | 41.4% | 8.3% |
| AA-Briefcase | 1822 | – |
Data as of 2026-09-22. Best configuration of each model; every score links to its source on the model pages.
Claude Opus 5.5 vs GPT-5.4: questions
- Is Claude Opus 5.5 better than GPT-5.4?
- Claude Opus 5.5 (thinking) leads on quality: 70.7 vs 65.1. The BenchLeader Index combines every independent quality benchmark; Claude Opus 5.5 (thinking) is ahead overall as of 2026-09-22, but check the category scores for your use.
- Which is cheaper, Claude Opus 5.5 or GPT-5.4?
- GPT-5.4 is cheaper: $5.63 against $8.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Opus 5.5 or GPT-5.4?
- GPT-5.4 streams faster: 140 against 48 output tokens per second.
- Which has the larger context window?
- GPT-5.4 accepts more context: 1.1M against 1M tokens.