Claude Sonnet 5.5 vs GPT-5.6 Terra
Claude Sonnet 5.5 
GPT-5.6 Terra Verdict
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 63.7.
- Claude Sonnet 5.5 (max) is stronger in agents & tools, coding, composite, knowledge, maths, reasoning.
- GPT-5.6 Terra (max) is stronger in long context, instruction following, multimodal.
- They cost about the same ($4.00 per 1M blended).
- Claude Sonnet 5.5 (max) streams 1.3× faster (141 vs 108 tokens per second).
| Metric | Claude Sonnet 5.5 (max) | GPT-5.6 Terra (max) |
|---|---|---|
| BenchLeader Index | 67.6 | 63.7 |
| Agents & tools score | 69.5 | 58.8 |
| Coding score | 69.8 | 64.7 |
| Composite score | 72.6 | 68.1 |
| Knowledge score | 65.4 | 57.5 |
| Long context score | 66.6 | 66.8 |
| Maths score | 72.9 | 69.3 |
| Reasoning score | 81.1 | 70.4 |
| Instruction following score | – | 69.8 |
| Multimodal score | – | 63.9 |
| Blended price $/M | $4.00 | $4.50 |
| Output speed | 141 tok/s | 108 tok/s |
| Time to first answer | 487.1 s | 113.7 s |
| Context window | 1M | 1.1M |
| GPQA Diamond | 95.6% | 93.3% |
| FrontierMath Tiers 1–3 | 88.8% | 86.0% |
| FrontierMath Tier 4 | 80.5% | 70.7% |
| OTIS Mock AIME | 100.0% | 99.7% |
| SimpleQA Verified | 46.5% | 43.2% |
| Terminal-Bench | 61.8% | 21.5% |
| SciCode | 61.0% | 55.0% |
| APEX-Agents | 75.5% | 58.2% |
| ProofBench | 100.0% | – |
| LMArena Agent | 12 | – |
| LiveBench | 75.7% | 77.9% |
| LiveBench Reasoning | 91.6% | 90.6% |
| LiveBench Coding | 91.4% | 78.3% |
| LiveBench Agentic Coding | 56.3% | 55.0% |
| LiveBench Mathematics | 96.1% | 94.9% |
| LiveBench Data Analysis | 59.5% | 79.3% |
| LiveBench Language | 78.0% | 82.9% |
| LiveBench Instruction Following | 56.8% | 64.6% |
| AA Intelligence Index v4.3.2 | 56 | 42.1 |
| IFBench | – | 71.2% |
| AA-LCR | 82.7% | 83.0% |
| MMMU-Pro | – | 80.7% |
| AA-Omniscience | 32.3 | 0.1 |
| Terminal-Bench Hard | – | 57.6% |
| GPQA Diamond (AA) | – | 92.5% |
| Humanity's Last Exam (AA) | 55.0% | 42.9% |
| SciCode (AA) | 61.0% | 55.0% |
| τ²-Bench Telecom (AA) | – | 86.3% |
| IOI | – | 87.6% |
| LegalBench | – | 85.1% |
| Terminal-Bench 2.1 (Vals) | – | 77.5% |
| SWE-bench (Vals) | – | 95.4% |
| Vals Index | – | 53.1 |
| ARC-AGI-1 | – | 96.5% |
| ARC-AGI-2 | – | 83.9% |
| ARC-AGI-3 | – | 0.8% |
| CritPt | 31.4% | 30.0% |
| GDPval-AA v2.1 | 67.0% | 47.7% |
| τ³-Banking (AA) | – | 40.2% |
| ITBench SRE (AA) | – | 51.0% |
| Analyst Agent (AA) | 57.5% | – |
| APEX-Agents (AA) | – | 38.9% |
| Code Migration | – | 47.8% |
| CyberBench | – | 72.1% |
| Excel Modeling Benchmark | – | 66.2% |
| Finance Agent v2 | – | 54.4% |
| Harvey's Legal Agent Benchmark | – | 0.8% |
| Legal Research Bench | – | 41.4% |
| MMMU-Pro (Vals) | – | 86.5% |
| ProgramBench | – | 0.5% |
| Public Benefits Bench | – | 62.4% |
| SAGE | – | 47.0% |
| SkillsBench | – | 58.9% |
| Tax Agent Bench | – | 65.2% |
| Terminal-Bench 4.0 (Vals) | – | 22.7% |
| Terminal-Bench Science | – | 10.0% |
| Vibe Code Bench 1-100 | – | 14.8% |
| Vibe Code Bench v1.1 | – | 74.6% |
| Chess Puzzles | – | 54.0% |
| Mystery Game Puzzles | 65.0% | 35.0% |
| BALROG | – | 53.2% |
| DeepSWE v1.1 | – | 69.6% |
| LMCA | – | 55.0% |
| DTBench | – | 93.3% |
| CursorBench | 55.5% | 41.3% |
| ALE-Bench | – | 1951.4 |
| FrontierSWE | 61.9% | – |
| Terminal-Bench 4.0 (AA) | 63.6% | 35.4% |
| Terminal-Bench 2.1 (AA) | – | 88.0% |
| AutomationBench | 71.8% | – |
| GDP.pdf | 25.8% | – |
| MLCR | 75.0% | – |
| Harvey LAB | 2.8% | – |
| AA-Omniscience: accuracy | 54.0% | 46.8% |
| AA-Omniscience: non-hallucination | 53.0% | 12.1% |
| 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-5.6 Terra: questions
- Is Claude Sonnet 5.5 better than GPT-5.6 Terra?
- Claude Sonnet 5.5 (max) leads on quality: 67.6 vs 63.7. 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-5.6 Terra for coding?
- Claude Sonnet 5.5 scores higher in coding (70 vs 65 on the category index, where 50 is average).
- Is Claude Sonnet 5.5 better than GPT-5.6 Terra for agentic tasks?
- Claude Sonnet 5.5 scores higher in agentic tasks (70 vs 59 on the category index, where 50 is average).
- Which is cheaper, Claude Sonnet 5.5 or GPT-5.6 Terra?
- Claude Sonnet 5.5 is cheaper: $4.00 against $4.50 per million tokens, blended at three input tokens per output token.
- Which is faster, Claude Sonnet 5.5 or GPT-5.6 Terra?
- Claude Sonnet 5.5 streams faster: 141 against 108 output tokens per second.
- Which has the larger context window?
- GPT-5.6 Terra accepts more context: 1.1M against 1M tokens.