| Benchmark | GPT-5.2 | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 60.2 | 47.9 |
| AA Intelligence | 43.3 | 33.5 |
| AA-LCR | 79.3 | 70.3 |
| AA-Omniscience | 0.3 | -21.4 |
| AIME 2025 | 100 | 94.5 |
| AIME25 no tools | 98 | 94.5 |
| Artificial Analysis Coding Index | 48.7 | 34.8 |
| browsecomp | 65.8 | 60.2 |
| browsecomp_with_context_manager | 65.8 | 60.2 |
| browsecomp_zh | 76.1 | 62.3 |
| critpt | 11.6 | 2.6 |
| frontiermath_tier_4 | 18.8 | 0 |
| gdpval | 48.3 | 24.5 |
| GPQA (unspecified) | 92.4 | 74.2 |
| GPQA Diamond | 92.4 | 84.5 |
| HLE | 45.5 | 23.9 |
| HLE (with tools) | 45.5 | 44.9 |
| HMMT 2025 | 99.4 | 89.4 |
| hmmt_feb_2025 | 100 | 93.3 |
| hmmt_nov_2025 | 99.2 | 89.2 |
| IFBench | 75.4 | 68.1 |
| imo_answer_bench | 86.3 | 78.6 |
| LiveCodeBench | 89 | 79.2 |
| longbench_v2 | 54.5 | 45.1 |
| MMLU-Pro | 87 | 84.6 |
| OmniScience Accuracy | 44.3 | 30.9 |
| OmniScience Non-Hallucination | 38.4 | 25.8 |
| scicode | 52.1 | 44.8 |
| Seal-0 | 45 | 56.3 |
| simplebench | 45.8 | 39.6 |
| swe_bench_bash | 72.8 | 63.4 |
| SWE-bench Multilingual | 72 | 61.1 |
| SWE-bench Verified | 80 | 71.3 |
| Terminal-Bench 2.0 | 54 | 35.7 |
| Terminal-Bench Hard | 62.2 | 31.1 |
| vectara_answer_rate | 100 | 98.6 |
| vectara_avg_summary_length | 186.3 | 59.2 |
| vectara_factual_consistency | 89.2 | 82.1 |
| vectara_hallucination_rate ↓ | 10.8 | 17.9 |
| τ²-Bench | 85.5 | 74.3 |
| τ²-Bench Telecom (AA run) | 98.7 | 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.