Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-R1-0528 | GPT-5.4 mini |
|---|---|---|
| GPQA Diamond | 81 | 88 |
| Humanity's Last Exam | 17.7 | 28.2 |
math
| Benchmark | DeepSeek-R1-0528 | GPT-5.4 mini |
|---|---|---|
| AIME 2025 | 87.5 | — |
coding
| Benchmark | DeepSeek-R1-0528 | GPT-5.4 mini |
|---|---|---|
| SWE-bench Verified | 44.6 | — |
| TerminalBench | 5.7 | — |
| SWE-Bench Pro | — | 54.4 |
agent
| Benchmark | DeepSeek-R1-0528 | GPT-5.4 mini |
|---|---|---|
| BrowseComp | 8.9 | — |
| Toolathlon | — | 42.9 |
| MCP Atlas | — | 57.7 |
multimodal
| Benchmark | DeepSeek-R1-0528 | GPT-5.4 mini |
|---|---|---|
| MMMU-Pro | — | 76.6 |
Category Winners
coding
35.8
math
65.6
reasoning
63.0
general
98.3
multimodal
72.7
agent
19.1