Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek R1 Distill Llama 70B | Nemotron 3 Super (120B A12B) |
|---|---|---|
| GPQA Diamond | 65.2 | 82.7 |
| Humanity's Last Exam | — | 22.8 |
math
| Benchmark | DeepSeek R1 Distill Llama 70B | Nemotron 3 Super (120B A12B) |
|---|---|---|
| AIME 2025 | — | 90.2 |
coding
| Benchmark | DeepSeek R1 Distill Llama 70B | Nemotron 3 Super (120B A12B) |
|---|---|---|
| SWE-bench Verified | — | 53.7 |
| TerminalBench | — | 25.8 |
| SciCode | — | 42.0 |
agent
| Benchmark | DeepSeek R1 Distill Llama 70B | Nemotron 3 Super (120B A12B) |
|---|---|---|
| BrowseComp | — | 31.3 |
Category Winners
coding
27.2
math
70.3
reasoning
53.7
agent
4.3