| Benchmark | Claude 4 Opus | Qwen3 VL 4B (Reasoning) |
|---|---|---|
| AA Intelligence | 31.7 | 7.7 |
| AA-LCR | 40 | 23 |
| AIME 2024 | 76 | 72.9 |
| AIME 2025 | 75.5 | 74.5 |
| Artificial Analysis Coding Index | 34 | 6.7 |
| GPQA Diamond | 79.6 | 64.1 |
| HLE | 12.3 | 4.6 |
| HMMT 2025 | 15.9 | 53.1 |
| IFBench | 53.7 | 36.6 |
| ifeval | 87.4 | 82.6 |
| LiveCodeBench | 70.4 | 51.3 |
| LiveCodeBench v6 | 47.4 | 51.3 |
| mmlu_redux | 94.2 | 86 |
| MMLU-Pro | 86.6 | 73.6 |
| scicode | 40.9 | 17.1 |
| supergpqa | 56.5 | 46.8 |
| Terminal-Bench Hard | 31.1 | 1.5 |
| τ²-Bench Telecom (AA run) | 73.4 | 15.5 |
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.