Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| GPQA Diamond | 92.4 | — |
| Humanity's Last Exam | 34.5 | — |
| ARC-AGI v2 | 52.9 | — |
math
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| AIME 2025 | 100 | 97.3 |
| FrontierMath | 40.3 | — |
coding
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| SWE-bench Verified | 80 | 74.4 |
multimodal
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| MMMU-Pro | 79.5 | — |
| CharXiv Reasoning | 82.1 | — |
| ScreenSpot Pro | 86.3 | — |
general
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| MMMLU | 89.6 | — |
agent
| Benchmark | GPT-5.2 | Step-3.5-Flash |
|---|---|---|
| BrowseComp | 65.8 | 69 |
| Toolathlon | 46.3 | — |
| MCP Atlas | 60.6 | — |
Category Winners
coding
81.1
math
S
Step-3.5-Flash81.0
reasoning
68.4
general
74.6
multimodal
80.3
agent
S
Step-3.5-Flash57.5