| Benchmark | DeepSeek-V3.2 | Muse Spark |
|---|---|---|
| AA Agentic Index | 39.8 | 28.7 |
| AA Intelligence | 32.8 | 44.3 |
| AA-LCR | 70.7 | 77 |
| AA-Omniscience | -22.5 | 7.2 |
| ARC-AGI-2 | 4 | 42.5 |
| arena_text_factuality | 1433 | 1467 |
| Artificial Analysis Coding Index | 44.2 | 58.6 |
| coding_arena_elo | 1368 | 1540 |
| critpt | 2.9 | 11.3 |
| cybergym | 17.3 | 43.5 |
| deepsearchqa_f1 | 60.9 | 74.8 |
| frontiermath_tier_4 | 2.1 | 14.6 |
| gdpval | 18.8 | 32.3 |
| GPQA Diamond | 84 | 89.5 |
| HLE | 40.8 | 58.4 |
| IFBench | 61 | 75.9 |
| MCP Atlas | 62.2 | 82.2 |
| OmniScience Accuracy | 33 | 49.6 |
| OmniScience Non-Hallucination | 17.3 | 26.8 |
| scicode | 39 | 51.5 |
| simpleqa_verified | 27.5 | 66.3 |
| SWE-bench Pro | 15.6 | 55 |
| SWE-bench Verified | 73.1 | 77.4 |
| TauBench V3 - Banking | 18.8 | 19.6 |
| Terminal-Bench 2.0 | 46.4 | 59 |
| Terminal-Bench 2.1 | 46.8 | 62.2 |
| Terminal-Bench Hard | 35.6 | 45.5 |
| τ²-Bench Telecom (AA run) | 90.6 | 91.5 |
| τ³-Bench | 69.2 | 19.6 |
Best tracked score per model per benchmark (default configuration; source-attributed). ↓ marks lower-is-better metrics. Open either model for its full surface, provenance and pricing.