Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
math
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3.6-27B |
|---|---|---|
| AIME 2025 | 96 | — |
reasoning
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3.6-27B |
|---|---|---|
| Humanity's Last Exam | 30.6 | 24 |
| GPQA Diamond | — | 87.8 |
coding
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3.6-27B |
|---|---|---|
| SWE-bench Verified | 73.1 | 77.2 |
| SWE-Bench Pro | — | 53.5 |
agent
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3.6-27B |
|---|---|---|
| Toolathlon | 35.2 | — |
multimodal
| Benchmark | DeepSeek-V3.2-Speciale | Qwen3.6-27B |
|---|---|---|
| MMMU | — | 82.9 |
| MMMU-Pro | — | 75.8 |
| CharXiv Reasoning | — | 78.4 |
Category Winners
coding
67.3
math
79.4
reasoning
Q
Qwen3.6-27B58.5
multimodal
Q
Qwen3.6-27B71.4
agent
11.2