Claude Fable 5 vs GPT-5.6 Terra
Claude Fable 5 
GPT-5.6 Terra Verdict
- Claude Fable 5 leads on quality: 70.4 vs 64.2.
- Claude Fable 5 is stronger in agents & tools, coding, composite, human preference, knowledge, long context, multimodal, reasoning.
- GPT-5.6 Terra (xhigh) is stronger in instruction following, maths.
- GPT-5.6 Terra (xhigh) is 4.4× cheaper ($4.50 vs $20.00 per 1M blended).
- GPT-5.6 Terra (xhigh) streams 1.5× faster (93 vs 63 tokens per second).
| Metric | Claude Fable 5 | GPT-5.6 Terra (xhigh) |
|---|---|---|
| BenchLeader Index | 70.4 | 64.2 |
| Agents & tools score | 74.4 | 69.6 |
| Coding score | 70.6 | 63.7 |
| Composite score | 91.6 | 77.1 |
| Human preference score | 71.3 | 66.3 |
| Instruction following score | 62.6 | 65.0 |
| Knowledge score | 71.3 | 60.9 |
| Long context score | 67.8 | 66.0 |
| Multimodal score | 69.6 | 63.0 |
| Reasoning score | 75.6 | 58.2 |
| Maths score | – | 71.0 |
| Blended price $/M | $20.00 | $4.50 |
| Output speed | 63 tok/s | 93 tok/s |
| Time to first answer | 92.2 s | 32.5 s |
| Context window | 1M | 1M |
| SimpleBench | – | 48.9% |
| SciCode | – | 51.6% |
| Remote Labor Index | 16.1% | – |
| APEX-Agents | 45.0% | – |
| FrontierCode | 53.5% | – |
| ProofBench | – | 74.0% |
| Epoch Capabilities Index | 163.4 | – |
| LMArena Text | 1507 | 1466 |
| LMArena Hard Prompts | 1533 | 1491 |
| LMArena Coding | 1551 | 1519 |
| LMArena WebDev | 1628 | 1521 |
| LMArena Vision | 1330 | 1269 |
| LMArena Agent | – | 1.5 |
| AA Intelligence Index | 49.7 | 38.2 |
| IFBench | 63.5% | 66.3% |
| AA-LCR | 82.3% | 79.0% |
| MMMU-Pro | – | 79.5% |
| AA-Omniscience | 43.3 | -3.0 |
| Terminal-Bench Hard | 62.9% | 62.9% |
| GPQA Diamond (AA) | 92.6% | 90.8% |
| Humanity's Last Exam (AA) | 55.5% | 41.9% |
| SciCode (AA) | 61.0% | 52.3% |
| τ²-Bench Telecom (AA) | 98.5% | 80.4% |
| LiveCodeBench | 89.8% | 85.9% |
| MMLU-Pro | 91.5% | 86.7% |
| IOI | 72.3% | 65.3% |
| LegalBench | 88.6% | – |
| CorpFin | 71.8% | 65.3% |
| TaxEval | 76.9% | 76.2% |
| Terminal-Bench 2.1 (Vals) | 80.5% | – |
| SWE-bench (Vals) | 95.0% | – |
| GPQA Diamond (Vals) | 93.2% | 90.9% |
| Vals Index | 66.0 | – |
| MCP Atlas | 83.3% | – |
| PRBench Finance | 53.9% | – |
| PRBench Legal | 52.6% | – |
| HiL-Bench | 56.3% | – |
| EQ-Bench 4 | 1340 | – |
| 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.