Claude Sonnet 5.5 vs GPT-6 Sol
Claude Sonnet 5.5 
GPT-6 Sol Verdict
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 65.6.
- Claude Sonnet 5.5 (max) is stronger in agents & tools, coding, maths, reasoning.
- GPT-6 Sol (max) is stronger in composite, knowledge, long context, human preference, multimodal.
- They cost about the same ($4.00 per 1M blended).
- Claude Sonnet 5.5 (max) streams 1.6× faster (141 vs 86 tokens per second).
| Metric | Claude Sonnet 5.5 (max) | GPT-6 Sol (max) |
|---|---|---|
| BenchLeader Index | 67.6 | 65.6 |
| Agents & tools score | 69.5 | 60.3 |
| Coding score | 69.8 | 67.6 |
| Composite score | 72.6 | 73.1 |
| Knowledge score | 65.4 | 70.1 |
| Long context score | 66.6 | 67.1 |
| Maths score | 72.9 | 69.5 |
| Reasoning score | 81.1 | 71.8 |
| Human preference score | – | 64.7 |
| Multimodal score | – | 61.9 |
| Blended price $/M | $4.00 | $4.00 |
| Output speed | 141 tok/s | 86 tok/s |
| Time to first answer | 487.1 s | 130.4 s |
| Context window | 1M | 1.1M |
| GPQA Diamond | 95.6% | 94.3% |
| FrontierMath Tiers 1–3 | 88.8% | 89.8% |
| FrontierMath Tier 4 | 80.5% | 90.0% |
| OTIS Mock AIME | 100.0% | 100.0% |
| SimpleQA Verified | 46.5% | 60.7% |
| Terminal-Bench | 61.8% | 49.4% |
| SciCode | 61.0% | 57.6% |
| APEX-Agents | 75.5% | 54.3% |
| FrontierCode | – | 49.3% |
| ProofBench | 100.0% | – |
| LMArena Text | – | 1456 |
| LMArena Hard Prompts | – | 1482 |
| LMArena Coding | – | 1519 |
| LMArena WebDev | – | 1688 |
| LMArena Vision | – | 1245 |
| LMArena Agent | 12 | 9.9 |
| LiveBench | 75.7% | 79.3% |
| LiveBench Reasoning | 91.6% | 88.7% |
| LiveBench Coding | 91.4% | 81.8% |
| LiveBench Agentic Coding | 56.3% | 52.9% |
| LiveBench Mathematics | 96.1% | 96.4% |
| LiveBench Data Analysis | 59.5% | 81.2% |
| LiveBench Language | 78.0% | 85.3% |
| LiveBench Instruction Following | 56.8% | 68.6% |
| AA Intelligence Index v4.3.2 | 56 | 47.6 |
| AA-LCR | 82.7% | 83.7% |
| MMMU-Pro | – | 83.0% |
| AA-Omniscience | 32.3 | 27.1 |
| Humanity's Last Exam (AA) | 55.0% | 47.9% |
| SciCode (AA) | 61.0% | 57.6% |
| IOI | – | 82.6% |
| Terminal-Bench 2.1 (Vals) | – | 83.2% |
| Vals Index | – | 57.5 |
| ARC-AGI-1 | – | 95.5% |
| ARC-AGI-2 | – | 89.6% |
| ARC-AGI-3 | – | 23.0% |
| CritPt | 31.4% | 30.9% |
| GDPval-AA v2.1 | 67.0% | 50.4% |
| ITBench SRE (AA) | – | 49.4% |
| Analyst Agent (AA) | 57.5% | – |
| BioMysteryBench | – | 74.8% |
| Code Migration | – | 57.2% |
| CyberBench | – | 78.0% |
| Excel Modeling Benchmark | – | 71.5% |
| Finance Agent v2 | – | 49.0% |
| Harvey's Legal Agent Benchmark | – | 1.7% |
| Legal Research Bench | – | 28.9% |
| MedCode | – | 47.1% |
| MedScribe | – | 82.0% |
| MysteryMechanism | – | 30.2% |
| ProgramBench | – | 2.0% |
| Public Benefits Bench | – | 56.6% |
| SAGE | – | 44.8% |
| Tax Agent Bench | – | 53.0% |
| Terminal-Bench 4.0 (Vals) | – | 44.4% |
| Terminal-Bench Science | – | 30.0% |
| Vibe Code Bench v1.1 | – | 87.8% |
| LMArena Maths | – | 1446 |
| LMArena Creative Writing | – | 1436 |
| LMArena Instruction Following | – | 1451 |
| LMArena Multi-turn | – | 1467 |
| LMArena Longer Queries | – | 1463 |
| EBR-bench | – | 53.3% |
| Mystery Game Puzzles | 65.0% | 56.0% |
| LMCA | – | 59.1% |
| DTBench | – | 97.3% |
| CursorBench | 55.5% | – |
| ALE-Bench | – | 2462 |
| GDP.pdf | – | 26.4% |
| FrontierSWE | 61.9% | – |
| Terminal-Bench 4.0 (AA) | 63.6% | 43.9% |
| AutomationBench | 71.8% | 61.6% |
| GDP.pdf | 25.8% | 25.2% |
| MLCR | 75.0% | 16.1% |
| Harvey LAB | 2.8% | 3.6% |
| AA-Omniscience: accuracy | 54.0% | 54.5% |
| AA-Omniscience: non-hallucination | 53.0% | 39.9% |
| AA-Briefcase v1.1 | 1823 | 1480 |
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-6 Sol: questions
- Is Claude Sonnet 5.5 better than GPT-6 Sol?
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 65.6. 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.
- Is Claude Sonnet 5.5 better than GPT-6 Sol for coding?
- Claude Sonnet 5.5 scores higher in coding (70 vs 68 on the category index, where 50 is average).
- Is Claude Sonnet 5.5 better than GPT-6 Sol for agentic tasks?
- Claude Sonnet 5.5 scores higher in agentic tasks (70 vs 60 on the category index, where 50 is average).
- Which is cheaper, Claude Sonnet 5.5 or GPT-6 Sol?
- Claude Sonnet 5.5 is cheaper: $4.00 against $4.00 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Sonnet 5.5 or GPT-6 Sol?
- Claude Sonnet 5.5 streams faster: 141 against 86 output tokens per second.
- Which has the larger context window?
- GPT-6 Sol accepts more context: 1.1M against 1M tokens.