| Benchmark | Claude Opus 4.8 | Qwen3.6 35B A3B |
|---|---|---|
| AA Agentic Index | 49.4 | 21.6 |
| AA Intelligence | 57.3 | 32.1 |
| AA-LCR | 73 | 66.7 |
| AA-Omniscience | 28.8 | -22.2 |
| aime_2026 | 100 | 92.7 |
| Artificial Analysis Coding Index | 74.3 | 41.9 |
| browsecomp | 88.5 | 48.7 |
| charxiv_rq | 80.5 | 78 |
| critpt | 20.9 | 0.3 |
| gdpval | 54.2 | 27.8 |
| GDPval-AA v2 | 1593 | 1015 |
| GPQA Diamond | 93.6 | 86 |
| HLE | 57.9 | 22.2 |
| hmmt_feb_2026 | 96.7 | 83.6 |
| hmmt_nov_2025 | 96.5 | 89.1 |
| IFBench | 62.2 | 64.4 |
| imo_answer_bench | 83.5 | 78.9 |
| MCP Atlas | 83.6 | 62.8 |
| MMMU-Pro | 78.9 | 75.3 |
| nl2repo | 69.7 | 29.4 |
| OmniScience Accuracy | 48.8 | 18.8 |
| OmniScience Non-Hallucination | 60.7 | 49.5 |
| scicode | 53.5 | 35.8 |
| SWE-bench Multilingual | 84.4 | 67.2 |
| SWE-bench Pro | 69.2 | 49.5 |
| SWE-bench Verified | 88.6 | 73.4 |
| TauBench V3 - Banking | 34.2 | 9.3 |
| Terminal-Bench 2.0 | 74.6 | 51.5 |
| Terminal-Bench 2.1 | 85 | 44.9 |
| Terminal-Bench Hard | 58.3 | 34.8 |
| tool_decathlon | 59.9 | 26.9 |
| toolathlon | 59.9 | 26.9 |
| WideSearch | 72.9 | 60.1 |
| τ²-Bench Telecom (AA run) | 94.4 | 95.3 |
| τ³-Bench | 27.6 | 67.2 |
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.