Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| Humanity's Last Exam | 30.6 | — |
| ARC-Challenge | — | 96.3 |
| GPQA Diamond | — | 49.3 |
| BigBench-Hard | — | 85.1 |
coding
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| SWE-bench Verified | 73.1 | — |
| HumanEval | — | 87.6 |
agent
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| Toolathlon | 35.2 | — |
language
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| WinoGrande | — | 87.5 |
| HellaSwag | — | 95.3 |
safety
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| TruthfulQA | — | 70.1 |
general
| Benchmark | DeepSeek-V3.2-Speciale | GPT-4 Turbo |
|---|---|---|
| Chatbot Arena Elo | — | 1256 |
| MMLU-Pro | — | 64.1 |
| IFEval | — | 82.1 |
Category Winners
coding
68.0
math
79.0
reasoning
50.4
general
47.9
language
85.2
safety
84.8
agent
10.2