Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| GPQA Diamond | 88.1 | 49.3 |
| Humanity's Last Exam | 45.1 | — |
| BigBench-Hard | — | 85.1 |
| ARC-Challenge | — | 96.3 |
coding
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| SWE-bench Verified | 79 | — |
| SWE-Bench Pro | 52.6 | — |
| HumanEval | — | 87.6 |
general
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| SimpleQA | 34.1 | — |
| MMLU-Pro | — | 64.1 |
| Chatbot Arena Elo | — | 1256 |
| IFEval | — | 82.1 |
agent
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| BrowseComp | 73.2 | — |
| Toolathlon | 47.8 | — |
| MCP Atlas | 69 | — |
language
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| HellaSwag | — | 95.3 |
| WinoGrande | — | 87.5 |
safety
| Benchmark | DeepSeek-V4-Flash-Max | GPT-4 Turbo |
|---|---|---|
| TruthfulQA | — | 70.1 |
Category Winners
coding
63.4
math
60.3
reasoning
80.9
general
47.9
language
85.2
safety
84.8
agent
48.4