Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT-5.4 | MiniMax M3 |
|---|---|---|
| GPQA Diamond | 92.8 | — |
| Humanity's Last Exam | 39.8 | — |
| ARC-AGI v2 | 73.3 | — |
coding
| Benchmark | GPT-5.4 | MiniMax M3 |
|---|---|---|
| SWE-Bench Pro | 57.7 | 59 |
| SWE-bench Verified | — | 80.5 |
multimodal
| Benchmark | GPT-5.4 | MiniMax M3 |
|---|---|---|
| MMMU-Pro | 81.2 | 78.1 |
agent
| Benchmark | GPT-5.4 | MiniMax M3 |
|---|---|---|
| BrowseComp | 82.7 | 83.5 |
| Toolathlon | 54.6 | — |
| MCP Atlas | 67.2 | 74.2 |
| APEX Agents | — | 27.7 |
math
| Benchmark | GPT-5.4 | MiniMax M3 |
|---|---|---|
| FrontierMath | 47.6 | — |
Category Winners
coding
68.3
math
69.1
reasoning
82.2
multimodal
83.5
agent
64.4