Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT OSS 20B | Kimi K2 Instruct |
|---|---|---|
| GPQA Diamond | 71.5 | 75.1 |
| Humanity's Last Exam | 10.9 | 4.7 |
agent
| Benchmark | GPT OSS 20B | Kimi K2 Instruct |
|---|---|---|
| TAU-Bench Retail | 54.8 | — |
math
| Benchmark | GPT OSS 20B | Kimi K2 Instruct |
|---|---|---|
| AIME 2025 | — | 49.5 |
general
| Benchmark | GPT OSS 20B | Kimi K2 Instruct |
|---|---|---|
| SimpleQA | — | 31 |
coding
| Benchmark | GPT OSS 20B | Kimi K2 Instruct |
|---|---|---|
| TerminalBench | — | 30 |
Category Winners
coding
M
Kimi K2 Instruct32.5
math
M
Kimi K2 Instruct7.4
reasoning
M
Kimi K2 Instruct35.1
general
M
Kimi K2 Instruct39.3
agent
16.1