Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Gemini 3.5 Flash | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| Humanity's Last Exam | 40.2 | — |
| ARC-AGI v2 | 72.1 | — |
coding
| Benchmark | Gemini 3.5 Flash | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| SWE-Bench Pro | 55.1 | — |
multimodal
| Benchmark | Gemini 3.5 Flash | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| MMMU-Pro | 83.6 | 68.1 |
| CharXiv Reasoning | 84.2 | 62.1 |
| ScreenSpot Pro | — | 62 |
agent
| Benchmark | Gemini 3.5 Flash | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| Toolathlon | 56.5 | — |
| MCP Atlas | 83.6 | — |
| OSWorld | — | 66.7 |
math
| Benchmark | Gemini 3.5 Flash | Qwen3 VL 235B A22B Instruct |
|---|---|---|
| AIME 2025 | — | 74.7 |
Category Winners
coding
32.7
math
Q
Qwen3 VL 235B A22B Instruct41.7
reasoning
78.0
general
Q
Qwen3 VL 235B A22B Instruct70.2
multimodal
83.0
agent
Q
Qwen3 VL 235B A22B Instruct83.0