| Benchmark | DeepSeek-R1 | GPT-5.5 |
|---|---|---|
| AA Agentic Index | 3.1 | 47.4 |
| AA Intelligence | 18.6 | 56.3 |
| AA-LCR | 56 | 79 |
| AA-Omniscience | -31.3 | 20.5 |
| arc_agi_1 | 21.2 | 95 |
| ARC-AGI-2 | 1.3 | 85 |
| Artificial Analysis Coding Index | 24.6 | 74.9 |
| browsecomp | 8.9 | 84.4 |
| critpt | 0.6 | 27.1 |
| Fortress | 74.4 | 16.3 |
| gdpval | 1.5 | 49.5 |
| GPQA Diamond | 81 | 93.6 |
| HLE | 17.7 | 52.2 |
| IFBench | 39 | 75.9 |
| OmniScience Accuracy | 30.7 | 58 |
| OmniScience Non-Hallucination | 10.5 | 12 |
| scicode | 35.7 | 56.1 |
| simplebench | 40.8 | 69 |
| simpleqa_verified | 27.4 | 63.1 |
| SWE-bench Verified | 57.6 | 80.6 |
| TauBench V3 - Banking | 6.4 | 39 |
| Terminal-Bench 2.1 | 19.1 | 84.3 |
| Terminal-Bench Hard | 6.1 | 60.6 |
| vectara_answer_rate | 97 | 100 |
| vectara_avg_summary_length | 93.5 | 129.6 |
| vectara_factual_consistency | 88.7 | 90.7 |
| vectara_hallucination_rate ↓ | 11.3 | 9.3 |
| τ²-Bench Telecom (AA run) | 11.4 | 93.9 |
| τ³-Bench | 6.4 | 31.3 |
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.