Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| GPQA Diamond | 88 | 78 |
| Humanity's Last Exam | 28.2 | 12.5 |
coding
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| SWE-Bench Pro | 54.4 | — |
| SWE-bench Verified | — | 69.4 |
| TerminalBench | — | 46.3 |
| SciCode | — | 36 |
multimodal
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| MMMU-Pro | 76.6 | — |
agent
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| Toolathlon | 42.9 | — |
| MCP Atlas | 57.7 | — |
| BrowseComp | — | 44 |
general
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| MRCR v2 | 33.6 | — |
math
| Benchmark | GPT-5.4 mini | MiniMax M2 |
|---|---|---|
| AIME 2025 | — | 78 |
Category Winners
coding
53.1
math
47.9
reasoning
65.3
general
34.1
multimodal
73.8
agent
22.3