| Benchmark | DeepSeek-V4-Flash | Qwen3.6 Plus |
|---|---|---|
| AA Agentic Index | 61.3 | 29 |
| AA Intelligence | 51.8 | 40.5 |
| AA-LCR | 74.3 | 72.3 |
| AA-Omniscience | -14.3 | 2.6 |
| aime_2026 | 95.8 | 95.3 |
| arena_elo | 1439 | 1444 |
| arena_text_factuality | 1434 | 1438 |
| Artificial Analysis Coding Index | 69.1 | 54.5 |
| coding_arena_elo | 1431 | 1460 |
| critpt | 16.6 | 2.9 |
| gdpval | 52.9 | 32 |
| GPQA Diamond | 90.8 | 90.4 |
| HLE | 38.6 | 28.8 |
| HLE (with tools) | 45.1 | 50.6 |
| hmmt_feb_2026 | 94.8 | 87.8 |
| IFBench | 79.2 | 75.2 |
| imo_answer_bench | 88.4 | 83.8 |
| livebench_agentic_coding | 46.8 | 41.4 |
| livebench_coding | 75 | 78.2 |
| livebench_data_analysis | 79.3 | 69.9 |
| livebench_instruction_following | 65.5 | 58.3 |
| livebench_language | 79.2 | 75 |
| livebench_math | 86.8 | 83.7 |
| livebench_reasoning | 86.6 | 75.8 |
| MCP Atlas | 69 | 74.1 |
| MMLU-Pro | 86.4 | 88.5 |
| nl2repo | 54.2 | 37.9 |
| OmniScience Accuracy | 40.4 | 26.4 |
| OmniScience Non-Hallucination | 15.6 | 68 |
| scicode | 49.9 | 40.7 |
| simpleqa_verified | 45 | 49.1 |
| SWE-bench Multilingual | 73.3 | 73.8 |
| SWE-bench Pro | 52.6 | 56.6 |
| SWE-bench Verified | 79 | 78.8 |
| TauBench V3 - Banking | 39.4 | 20.8 |
| Terminal-Bench 2.0 | 56.9 | 61.6 |
| Terminal-Bench 2.1 | 82.7 | 61.4 |
| Terminal-Bench Hard | 38.6 | 43.9 |
| toolathlon | 70.3 | 39.8 |
| τ²-Bench Telecom (AA run) | 95.6 | 97.7 |
| τ³-Bench | 22.9 | 70.7 |
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.