Claude Sonnet 5.5 vs GPT-5.4 Pro
Claude Sonnet 5.5 
GPT-5.4 Pro Verdict
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 63.6.
- Claude Sonnet 5.5 (max) is stronger in agents & tools, coding, composite, long context, maths, reasoning.
- GPT-5.4 Pro is stronger in knowledge, instruction following, multimodal.
- Claude Sonnet 5.5 (max) is 17× cheaper ($4.00 vs $67.50 per 1M blended).
- Claude Sonnet 5.5 (max) streams 23.5× faster (141 vs 6 tokens per second).
| Metric | Claude Sonnet 5.5 (max) | GPT-5.4 Pro |
|---|---|---|
| BenchLeader Index | 67.6 | 63.6 |
| Agents & tools score | 69.5 | – |
| Coding score | 69.8 | – |
| Composite score | 72.6 | – |
| Knowledge score | 65.4 | 73.6 |
| Long context score | 66.6 | – |
| Maths score | 72.9 | – |
| Reasoning score | 81.1 | 73.2 |
| Instruction following score | – | 65.3 |
| Multimodal score | – | 69.8 |
| Blended price $/M | $4.00 | $67.50 |
| Output speed | 141 tok/s | 6 tok/s |
| Time to first answer | 487.1 s | 7.0 s |
| Context window | 1M | 1.1M |
| GPQA Diamond | 95.6% | – |
| FrontierMath Tiers 1–3 | 88.8% | – |
| FrontierMath Tier 4 | 80.5% | – |
| OTIS Mock AIME | 100.0% | – |
| SimpleQA Verified | 46.5% | – |
| Humanity's Last Exam | – | 44.3% |
| Terminal-Bench | 61.8% | – |
| SimpleBench | – | 74.1% |
| SciCode | 61.0% | – |
| APEX-Agents | 75.5% | – |
| ProofBench | 100.0% | – |
| Epoch Capabilities Index | – | 158.9 |
| LMArena Agent | 12 | – |
| LiveBench | 75.7% | – |
| LiveBench Reasoning | 91.6% | – |
| LiveBench Coding | 91.4% | – |
| LiveBench Agentic Coding | 56.3% | – |
| LiveBench Mathematics | 96.1% | – |
| LiveBench Data Analysis | 59.5% | – |
| LiveBench Language | 78.0% | – |
| LiveBench Instruction Following | 56.8% | – |
| AA Intelligence Index v4.3.2 | 56 | – |
| AA-LCR | 82.7% | – |
| AA-Omniscience | 32.3 | – |
| Humanity's Last Exam (AA) | 55.0% | – |
| SciCode (AA) | 61.0% | – |
| MultiChallenge | – | 69.2% |
| VISTA | – | 53.9% |
| MultiNRC | – | 62.3% |
| TutorBench | – | 56.6% |
| CritPt | 31.4% | – |
| GDPval-AA v2.1 | 67.0% | – |
| Analyst Agent (AA) | 57.5% | – |
| FORTRESS | – | 14.8% |
| MASK | – | 91.7% |
| Mystery Game Puzzles | 65.0% | – |
| CursorBench | 55.5% | – |
| FrontierSWE | 61.9% | – |
| Terminal-Bench 4.0 (AA) | 63.6% | – |
| AutomationBench | 71.8% | – |
| GDP.pdf | 25.8% | – |
| MLCR | 75.0% | – |
| Harvey LAB | 2.8% | – |
| AA-Omniscience: accuracy | 54.0% | – |
| AA-Omniscience: non-hallucination | 53.0% | – |
| AA-Briefcase v1.1 | 1823 | – |
Data as of 2026-10-11. Best configuration of each model; every score links to its source on the model pages.
Claude Sonnet 5.5 vs GPT-5.4 Pro: questions
- Is Claude Sonnet 5.5 better than GPT-5.4 Pro?
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 63.6. The BenchLeader Index combines every independent quality benchmark; Claude Sonnet 5.5 (max) is ahead overall as of 2026-10-11, but check the category scores for your use.
- Which is cheaper, Claude Sonnet 5.5 or GPT-5.4 Pro?
- Claude Sonnet 5.5 is cheaper: $4.00 against $67.50 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Sonnet 5.5 or GPT-5.4 Pro?
- Claude Sonnet 5.5 streams faster: 141 against 6 output tokens per second.
- Which has the larger context window?
- GPT-5.4 Pro accepts more context: 1.1M against 1M tokens.