Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| ARC-Challenge | 96.7 | — |
| BigBench-Hard | 89.4 | — |
| GPQA Diamond | 65 | — |
| Humanity's Last Exam | — | 30.6 |
language
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| HellaSwag | 89 | — |
| WinoGrande | 83.2 | — |
coding
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| SWE-bench Verified | 49 | 73.1 |
| HumanEval | 92 | — |
multimodal
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| MMMU | 68.3 | — |
safety
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| TruthfulQA | 74.1 | — |
general
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| MMLU-Pro | 78 | — |
| Chatbot Arena Elo | 1271 | — |
| IFEval | 88 | — |
agent
| Benchmark | Claude 3.5 Sonnet | DeepSeek-V3.2-Speciale |
|---|---|---|
| Toolathlon | — | 35.2 |
Category Winners
coding
68.0
math
79.0
reasoning
68.4
general
80.5
language
31.6
multimodal
52.2
safety
95.3
agent