Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
coding
| Benchmark | GPT-5.2 Codex | Qwen3 VL 32B Instruct |
|---|---|---|
| SWE-Bench Pro | 56.4 | — |
reasoning
| Benchmark | GPT-5.2 Codex | Qwen3 VL 32B Instruct |
|---|---|---|
| GPQA Diamond | — | 68.9 |
math
| Benchmark | GPT-5.2 Codex | Qwen3 VL 32B Instruct |
|---|---|---|
| AIME 2025 | — | 66.2 |
multimodal
| Benchmark | GPT-5.2 Codex | Qwen3 VL 32B Instruct |
|---|---|---|
| MMMU-Pro | — | 65.3 |
| CharXiv Reasoning | — | 62.8 |
| ScreenSpot Pro | — | 57.9 |
agent
| Benchmark | GPT-5.2 Codex | Qwen3 VL 32B Instruct |
|---|---|---|
| OSWorld | — | 32.6 |
Category Winners
coding
39.8
math
Q
Qwen3 VL 32B Instruct26.9
reasoning
Q
Qwen3 VL 32B Instruct54.4
multimodal
Q
Qwen3 VL 32B Instruct35.1
agent
Q
Qwen3 VL 32B Instruct26.8