Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-V3.2 | GPT-5.4 |
|---|---|---|
| GPQA Diamond | 82.4 | 92.8 |
| Humanity's Last Exam | 40.8 | 39.8 |
| ARC-AGI v2 | — | 73.3 |
math
| Benchmark | DeepSeek-V3.2 | GPT-5.4 |
|---|---|---|
| AIME 2025 | 93.1 | — |
| FrontierMath | — | 47.6 |
coding
| Benchmark | DeepSeek-V3.2 | GPT-5.4 |
|---|---|---|
| SWE-bench Verified | 73.1 | — |
| SWE-Bench Pro | — | 57.7 |
agent
| Benchmark | DeepSeek-V3.2 | GPT-5.4 |
|---|---|---|
| BrowseComp | 51.4 | 82.7 |
| Toolathlon | 35.2 | 54.6 |
| MCP Atlas | — | 67.2 |
multimodal
| Benchmark | DeepSeek-V3.2 | GPT-5.4 |
|---|---|---|
| MMMU-Pro | — | 81.2 |
Category Winners
coding
68.0
math
74.8
reasoning
82.2
multimodal
83.5
agent
64.4