GPT-5.6 Terra vs Muse Spark
GPT-5.6 Terra 
Muse Spark Verdict
- Muse Spark leads on quality: 65.4 vs 64.2.
- GPT-5.6 Terra (xhigh) is stronger in agents & tools, composite, long context, maths.
- Muse Spark is stronger in coding, human preference, instruction following, knowledge, multimodal, reasoning.
| Metric | GPT-5.6 Terra (xhigh) | Muse Spark |
|---|---|---|
| BenchLeader Index | 64.2 | 65.4 |
| Agents & tools score | 69.6 | 66.3 |
| Coding score | 63.7 | 66.0 |
| Composite score | 77.1 | 68.4 |
| Human preference score | 66.3 | 69.0 |
| Instruction following score | 65.0 | 79.6 |
| Knowledge score | 60.9 | 64.5 |
| Long context score | 66.0 | 65.5 |
| Maths score | 71.0 | 56.3 |
| Multimodal score | 63.0 | 65.9 |
| Reasoning score | 58.2 | 70.0 |
| Blended price $/M | $4.50 | – |
| Output speed | 93 tok/s | – |
| Time to first answer | 32.5 s | – |
| Context window | 1M | 262k |
| GPQA Diamond | – | 89.8% |
| OTIS Mock AIME | – | 88.9% |
| Humanity's Last Exam | – | 40.6% |
| SimpleBench | 48.9% | – |
| SciCode | 51.6% | 51.5% |
| ProofBench | 74.0% | 17.0% |
| Epoch Capabilities Index | – | 152.1 |
| LMArena Text | 1466 | 1488 |
| LMArena Hard Prompts | 1491 | 1505 |
| LMArena Coding | 1519 | 1526 |
| LMArena WebDev | 1521 | – |
| LMArena Vision | 1269 | 1306 |
| LMArena Agent | 1.5 | – |
| AA Intelligence Index | 38.2 | 31.3 |
| IFBench | 66.3% | 75.9% |
| AA-LCR | 79.0% | 78.0% |
| MMMU-Pro | 79.5% | 80.5% |
| AA-Omniscience | -3.0 | 7.2 |
| Terminal-Bench Hard | 62.9% | 45.5% |
| GPQA Diamond (AA) | 90.8% | 88.4% |
| Humanity's Last Exam (AA) | 41.9% | 40.7% |
| SciCode (AA) | 52.3% | – |
| τ²-Bench Telecom (AA) | 80.4% | 91.5% |
| AIME (Vals) | – | 96.9% |
| LiveCodeBench | 85.9% | – |
| MMLU-Pro | 86.7% | 87.3% |
| IOI | 65.3% | – |
| LegalBench | – | 84.2% |
| CorpFin | 65.3% | 65.1% |
| TaxEval | 76.2% | 77.7% |
| SWE-bench (Vals) | – | 74.4% |
| GPQA Diamond (Vals) | 90.9% | 89.7% |
| SWE-Bench Pro | – | 55.0% |
| MCP Atlas | – | 82.2% |
| MultiChallenge | – | 75.5% |
| PRBench Finance | – | 52.4% |
| PRBench Legal | – | 52.3% |
| MultiNRC | – | 59.0% |
| TutorBench | – | 68.5% |
| 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.