Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Gemini 3 Flash | Qwen3.5-27B |
|---|---|---|
| GPQA Diamond | 90.4 | 85.5 |
| Humanity's Last Exam | 43.5 | 48.5 |
| ARC-AGI v2 | 33.6 | — |
math
| Benchmark | Gemini 3 Flash | Qwen3.5-27B |
|---|---|---|
| AIME 2025 | 99.7 | — |
coding
| Benchmark | Gemini 3 Flash | Qwen3.5-27B |
|---|---|---|
| SWE-bench Verified | 78 | 72.4 |
multimodal
| Benchmark | Gemini 3 Flash | Qwen3.5-27B |
|---|---|---|
| MMMU-Pro | 81.2 | 75 |
| CharXiv Reasoning | 80.3 | 79.5 |
| ScreenSpot Pro | 69.1 | 70.3 |
| MMMU | — | 82.3 |
agent
| Benchmark | Gemini 3 Flash | Qwen3.5-27B |
|---|---|---|
| Toolathlon | 49.4 | — |
| MCP Atlas | 57.4 | — |
| BrowseComp | — | 61 |
Category Winners
coding
77.7
math
83.7
reasoning
Q
Qwen3.5-27B83.0
general
62.6
multimodal
72.4
agent
Q
Qwen3.5-27B42.1