| Benchmark | GPT-5.4 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 58.2 | 47.9 |
| AA Intelligence | 53.1 | 33.5 |
| AA-LCR | 77.7 | 70.3 |
| AA-Omniscience | 5.8 | -21.4 |
| Artificial Analysis Coding Index | 71.1 | 34.8 |
| browsecomp | 82.7 | 60.2 |
| browsecomp_with_context_manager | 82.7 | 60.2 |
| critpt | 23.4 | 2.6 |
| frontiermath_tier_4 | 27.1 | 0 |
| gdpval | 50.1 | 24.5 |
| GPQA Diamond | 93 | 84.5 |
| HLE | 43.7 | 23.9 |
| HLE (with tools) | 52.1 | 44.9 |
| hmmt_nov_2025 | 95.8 | 89.2 |
| IFBench | 73.9 | 68.1 |
| imo_answer_bench | 91.4 | 78.6 |
| MMLU-Pro | 87.5 | 84.6 |
| OmniScience Accuracy | 50.9 | 30.9 |
| OmniScience Non-Hallucination | 17.4 | 25.8 |
| scicode | 56.6 | 44.8 |
| SWE-bench Multilingual | 71.7 | 61.1 |
| Terminal-Bench 2.0 | 75.1 | 35.7 |
| Terminal-Bench Hard | 57.6 | 31.1 |
| vectara_answer_rate | 99.9 | 98.6 |
| vectara_avg_summary_length | 81.7 | 59.2 |
| vectara_factual_consistency | 93 | 82.1 |
| vectara_hallucination_rate ↓ | 7 | 17.9 |
| τ²-Bench Telecom (AA run) | 87.1 | 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.