| Benchmark | GPT-5.5 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 47.4 | 47.9 |
| AA Intelligence | 56.3 | 33.5 |
| AA-LCR | 79 | 70.3 |
| AA-Omniscience | 20.5 | -21.4 |
| Artificial Analysis Coding Index | 74.9 | 34.8 |
| browsecomp | 84.4 | 60.2 |
| critpt | 27.1 | 2.6 |
| frontiermath_tier_4 | 35.4 | 0 |
| gdpval | 49.5 | 24.5 |
| GPQA Diamond | 93.6 | 84.5 |
| HealthBench | 56.5 | 58 |
| HLE | 52.2 | 23.9 |
| HLE (with tools) | 52.2 | 44.9 |
| hmmt_nov_2025 | 96.5 | 89.2 |
| IFBench | 75.9 | 68.1 |
| OmniScience Accuracy | 58 | 30.9 |
| OmniScience Non-Hallucination | 12 | 25.8 |
| ResearchRubrics | 64 | 56.2 |
| scicode | 56.1 | 44.8 |
| simplebench | 69 | 39.6 |
| SWE-bench Verified | 80.6 | 71.3 |
| Terminal-Bench 2.0 | 82.7 | 35.7 |
| Terminal-Bench Hard | 60.6 | 31.1 |
| vectara_answer_rate | 100 | 98.6 |
| vectara_avg_summary_length | 129.6 | 59.2 |
| vectara_factual_consistency | 90.7 | 82.1 |
| vectara_hallucination_rate ↓ | 9.3 | 17.9 |
| τ²-Bench Telecom (AA run) | 93.9 | 93 |
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.