| Benchmark | DeepSeek-R1 | Gemini 3.6 Flash |
|---|---|---|
| AA Agentic Index | 3.1 | 40.5 |
| AA Intelligence | 18.6 | 52 |
| AA-LCR | 56 | 79 |
| AA-Omniscience | -31.3 | 22.1 |
| arc_agi_1 | 21.2 | 34.5 |
| ARC-AGI-2 | 1.3 | 2.6 |
| Artificial Analysis Coding Index | 24.6 | 69.2 |
| critpt | 0.6 | 10.6 |
| gdpval | 1.5 | 46.1 |
| GPQA Diamond | 81 | 92.8 |
| HLE | 17.7 | 40.8 |
| OmniScience Accuracy | 30.7 | 50 |
| OmniScience Non-Hallucination | 10.5 | 44.4 |
| scicode | 35.7 | 52.7 |
| simpleqa_verified | 27.4 | 68.7 |
| TauBench V3 - Banking | 6.4 | 29.9 |
| Terminal-Bench 2.1 | 19.1 | 78 |
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.