Claude Fable 5.1 vs GPT-5.6 Terra
Claude Fable 5.1 
GPT-5.6 Terra Verdict
- Claude Fable 5.1 (xhigh) leads on quality: 70.3 vs 64.2.
- Claude Fable 5.1 (xhigh) is stronger in coding, composite, knowledge, long context, reasoning.
- GPT-5.6 Terra (xhigh) is stronger in agents & tools, human preference, instruction following, maths, multimodal.
- GPT-5.6 Terra (xhigh) is 4.4× cheaper ($4.50 vs $20.00 per 1M blended).
- GPT-5.6 Terra (xhigh) streams 1.6× faster (93 vs 58 tokens per second).
| Metric | Claude Fable 5.1 (xhigh) | GPT-5.6 Terra (xhigh) |
|---|---|---|
| BenchLeader Index | 70.3 | 64.2 |
| Coding score | 71.3 | 63.7 |
| Composite score | 95.0 | 77.1 |
| Knowledge score | 84.1 | 60.9 |
| Long context score | 68.1 | 66.0 |
| Reasoning score | 76.4 | 58.2 |
| Agents & tools score | – | 69.6 |
| Human preference score | – | 66.3 |
| Instruction following score | – | 65.0 |
| Maths score | – | 71.0 |
| Multimodal score | – | 63.0 |
| Blended price $/M | $20.00 | $4.50 |
| Output speed | 58 tok/s | 93 tok/s |
| Time to first answer | 132.0 s | 32.5 s |
| Context window | 1M | 1M |
| Humanity's Last Exam | 46.5% | – |
| SimpleBench | – | 48.9% |
| SciCode | 60.1% | 51.6% |
| ProofBench | – | 74.0% |
| LMArena Text | – | 1466 |
| LMArena Hard Prompts | – | 1491 |
| LMArena Coding | – | 1519 |
| LMArena WebDev | – | 1521 |
| LMArena Vision | – | 1269 |
| LMArena Agent | – | 1.5 |
| AA Intelligence Index | 53.2 | 38.2 |
| IFBench | – | 66.3% |
| AA-LCR | 83.0% | 79.0% |
| MMMU-Pro | – | 79.5% |
| AA-Omniscience | 42.4 | -3.0 |
| Terminal-Bench Hard | – | 62.9% |
| GPQA Diamond (AA) | 93.4% | 90.8% |
| Humanity's Last Exam (AA) | 58.7% | 41.9% |
| SciCode (AA) | 60.9% | 52.3% |
| τ²-Bench Telecom (AA) | – | 80.4% |
| LiveCodeBench | – | 85.9% |
| MMLU-Pro | – | 86.7% |
| IOI | – | 65.3% |
| CorpFin | – | 65.3% |
| TaxEval | – | 76.2% |
| GPQA Diamond (Vals) | – | 90.9% |
| ARC-AGI-1 | 96.5% | 94.0% |
| ARC-AGI-2 | 90.0% | 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.