| Benchmark | DeepSeek-R1 | GPT-5.6 Sol |
|---|---|---|
| AA Agentic Index | 3.1 | 57.8 |
| AA Intelligence | 18.6 | 61 |
| AA-LCR | 56 | 77.7 |
| AA-Omniscience | -31.3 | 22 |
| arc_agi_1 | 21.2 | 97.5 |
| ARC-AGI-2 | 1.3 | 92.5 |
| Artificial Analysis Coding Index | 24.6 | 78.3 |
| browsecomp | 8.9 | 90.4 |
| critpt | 0.6 | 32.3 |
| gdpval | 1.5 | 61.1 |
| GPQA Diamond | 81 | 94.6 |
| HLE | 17.7 | 49.5 |
| IFBench | 39 | 72.7 |
| longbench_v2 | 58.3 | 67.1 |
| OmniScience Accuracy | 30.7 | 59.4 |
| OmniScience Non-Hallucination | 10.5 | 10.6 |
| scicode | 35.7 | 56.9 |
| simplebench | 40.8 | 64.8 |
| simpleqa_verified | 27.4 | 71.6 |
| TauBench V3 - Banking | 6.4 | 44.3 |
| Terminal-Bench 2.1 | 19.1 | 89.5 |
| Terminal-Bench Hard | 6.1 | 65.9 |
| τ²-Bench Telecom (AA run) | 11.4 | 85.1 |
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.