Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
agent
| Benchmark | DeepSeek-V3.2-Speciale | Gemini 2.5 Pro Preview 06-05 |
|---|---|---|
| Toolathlon | 35.2 | — |
math
| Benchmark | DeepSeek-V3.2-Speciale | Gemini 2.5 Pro Preview 06-05 |
|---|---|---|
| AIME 2025 | 96 | 88 |
reasoning
| Benchmark | DeepSeek-V3.2-Speciale | Gemini 2.5 Pro Preview 06-05 |
|---|---|---|
| Humanity's Last Exam | 30.6 | 21.6 |
| GPQA Diamond | — | 86.4 |
coding
| Benchmark | DeepSeek-V3.2-Speciale | Gemini 2.5 Pro Preview 06-05 |
|---|---|---|
| SWE-bench Verified | 73.1 | 67.2 |
multimodal
| Benchmark | DeepSeek-V3.2-Speciale | Gemini 2.5 Pro Preview 06-05 |
|---|---|---|
| MMMU | — | 82 |
Category Winners
coding
67.7
math
79.2
reasoning
55.3
general
42.5
multimodal
83.8
agent
11.2