Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT-5 nano | Qwen3 VL 4B Thinking |
|---|---|---|
| GPQA Diamond | 71.2 | 64.1 |
| Humanity's Last Exam | 8.7 | — |
math
| Benchmark | GPT-5 nano | Qwen3 VL 4B Thinking |
|---|---|---|
| AIME 2025 | 85.2 | 74.5 |
| FrontierMath | 9.6 | — |
multimodal
| Benchmark | GPT-5 nano | Qwen3 VL 4B Thinking |
|---|---|---|
| MMMU-Pro | — | 57 |
| CharXiv Reasoning | — | 50.3 |
| ScreenSpot Pro | — | 49.2 |
agent
| Benchmark | GPT-5 nano | Qwen3 VL 4B Thinking |
|---|---|---|
| OSWorld | — | 31.4 |
Category Winners
math
Q
Qwen3 VL 4B Thinking41.2
reasoning
Q
Qwen3 VL 4B Thinking43.0
multimodal
Q
Qwen3 VL 4B Thinking14.0
agent
Q
Qwen3 VL 4B Thinking25.0