Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| GPQA Diamond | 83 | — |
| Humanity's Last Exam | 17.8 | — |
| ARC-AGI v2 | 4.9 | — |
math
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| AIME 2025 | 83 | 74.7 |
coding
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| SWE-bench Verified | 63.2 | — |
multimodal
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| MMMU | 79.6 | — |
| MMMU-Pro | — | 68.1 |
| CharXiv Reasoning | — | 62.1 |
| ScreenSpot Pro | — | 62 |
general
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| SimpleQA | 50.8 | 51.9 |
agent
| Benchmark | Gemini 2.5 Pro | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| OSWorld | — | 66.7 |
Category Winners
coding
43.5
math
57.2
reasoning
35.5
general
Q
Qwen3 VL 235B A22B Instruct70.6
multimodal
79.3
agent
Q
Qwen3 VL 235B A22B Instruct82.3