GPT-5.5 vs GPT-5.6 Terra
GPT-5.5 Verdict
- GPT-5.5 leads on quality: 66.6 vs 64.2.
- GPT-5.5 is stronger in composite, human preference, instruction following, knowledge, long context, multimodal, reasoning.
- GPT-5.6 Terra (xhigh) is stronger in agents & tools, coding, maths.
- GPT-5.6 Terra (xhigh) is 2.5× cheaper ($4.50 vs $11.25 per 1M blended).
| Metric | GPT-5.5 | GPT-5.6 Terra (xhigh) |
|---|---|---|
| BenchLeader Index | 66.6 | 64.2 |
| Agents & tools score | 68.4 | 69.6 |
| Coding score | 57.3 | 63.7 |
| Composite score | 77.7 | 77.1 |
| Human preference score | 67.6 | 66.3 |
| Instruction following score | 73.3 | 65.0 |
| Knowledge score | 73.7 | 60.9 |
| Long context score | 68.8 | 66.0 |
| Multimodal score | 64.9 | 63.0 |
| Reasoning score | 69.7 | 58.2 |
| Maths score | – | 71.0 |
| Blended price $/M | $11.25 | $4.50 |
| Output speed | 88 tok/s | 93 tok/s |
| Time to first answer | 62.3 s | 32.5 s |
| Context window | 1.1M | 1M |
| Terminal-Bench | 84.7% | – |
| SimpleBench | 69.0% | 48.9% |
| SciCode | – | 51.6% |
| Remote Labor Index | 6.3% | – |
| APEX-Agents | 38.5% | – |
| FrontierCode | 43.0% | – |
| ProofBench | – | 74.0% |
| Epoch Capabilities Index | 159.1 | – |
| LMArena Text | 1477 | 1466 |
| LMArena Hard Prompts | 1498 | 1491 |
| LMArena Coding | 1509 | 1519 |
| LMArena WebDev | 1458 | 1521 |
| LMArena Vision | 1296 | 1269 |
| LMArena Agent | 3 | 1.5 |
| AA Intelligence Index | 38.6 | 38.2 |
| IFBench | 75.8% | 66.3% |
| AA-LCR | 84.3% | 79.0% |
| MMMU-Pro | 79.9% | 79.5% |
| AA-Omniscience | 20.5 | -3.0 |
| Terminal-Bench Hard | 60.6% | 62.9% |
| GPQA Diamond (AA) | 93.5% | 90.8% |
| Humanity's Last Exam (AA) | 45.8% | 41.9% |
| SciCode (AA) | 55.8% | 52.3% |
| τ²-Bench Telecom (AA) | 93.9% | 80.4% |
| LiveCodeBench | – | 85.9% |
| MMLU-Pro | – | 86.7% |
| IOI | – | 65.3% |
| CorpFin | – | 65.3% |
| TaxEval | – | 76.2% |
| GPQA Diamond (Vals) | – | 90.9% |
| HiL-Bench | 39.7% | – |
| EQ-Bench 4 | 1315 | – |
| Kagi LLM Benchmark | 88.8% | – |
| ARC-AGI-1 | – | 94.0% |
| ARC-AGI-2 | – | 74.2% |
| ARC-AGI-3 | – | 0.7% |
Data as of 2026-09-09. Best configuration of each model; every score links to its source on the model pages.