| Benchmark | DeepSeek-V4-Flash | Kimi K2 (Reasoning) |
|---|---|---|
| AA Agentic Index | 61.3 | 47.9 |
| AA Intelligence | 51.8 | 33.5 |
| AA-LCR | 74.3 | 70.3 |
| AA-Omniscience | -14.3 | -21.4 |
| Artificial Analysis Coding Index | 69.1 | 34.8 |
| browsecomp | 73.2 | 60.2 |
| critpt | 16.6 | 2.6 |
| gdpval | 52.9 | 24.5 |
| GPQA Diamond | 90.8 | 84.5 |
| HLE | 38.6 | 23.9 |
| HLE (with tools) | 45.1 | 44.9 |
| IFBench | 79.2 | 68.1 |
| imo_answer_bench | 88.4 | 78.6 |
| LiveCodeBench | 91.6 | 79.2 |
| MMLU-Pro | 86.4 | 84.6 |
| OmniScience Accuracy | 40.4 | 30.9 |
| OmniScience Non-Hallucination | 15.6 | 25.8 |
| scicode | 49.9 | 44.8 |
| simplebench | 46.3 | 39.6 |
| SWE-bench Multilingual | 73.3 | 61.1 |
| SWE-bench Verified | 79 | 71.3 |
| Terminal-Bench 2.0 | 56.9 | 35.7 |
| Terminal-Bench Hard | 38.6 | 31.1 |
| τ²-Bench Telecom (AA run) | 95.6 | 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.