Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | DeepSeek-V3.2 | GPT OSS 120B High |
|---|---|---|
| GPQA Diamond | 82.4 | 80.9 |
| Humanity's Last Exam | 40.8 | — |
math
| Benchmark | DeepSeek-V3.2 | GPT OSS 120B High |
|---|---|---|
| AIME 2025 | 93.1 | 92.5 |
coding
| Benchmark | DeepSeek-V3.2 | GPT OSS 120B High |
|---|---|---|
| SWE-bench Verified | 73.1 | — |
agent
| Benchmark | DeepSeek-V3.2 | GPT OSS 120B High |
|---|---|---|
| BrowseComp | 51.4 | — |
| Toolathlon | 35.2 | — |
general
| Benchmark | DeepSeek-V3.2 | GPT OSS 120B High |
|---|---|---|
| MMMLU | — | 83.8 |
Category Winners
coding
68.0
math
74.8
reasoning
76.0
general
52.1
agent
17.6