Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-V3.1 | K2 Horizon 32B |
|---|---|---|
| GPQA Diamond | 74.9 | 82.8 |
| Humanity's Last Exam | 15.9 | 25.2 |
math
| Benchmark | DeepSeek-V3.1 | K2 Horizon 32B |
|---|---|---|
| AIME 2025 | 49.8 | — |
coding
| Benchmark | DeepSeek-V3.1 | K2 Horizon 32B |
|---|---|---|
| SWE-bench Verified | 66 | — |
| TerminalBench | 31.3 | — |
| SciCode | — | 37.9 |
general
| Benchmark | DeepSeek-V3.1 | K2 Horizon 32B |
|---|---|---|
| SimpleQA | 93.4 | — |
agent
| Benchmark | DeepSeek-V3.1 | K2 Horizon 32B |
|---|---|---|
| BrowseComp | 30 | — |
Category Winners
coding
43.8
math
8.0
reasoning
I
K2 Horizon 32B56.2
general
98.5
agent
3.2