Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | GPT-4o | Kimi K2 Instruct |
|---|---|---|
| GPQA Diamond | 70.1 | 75.1 |
| Humanity's Last Exam | 5.3 | 4.7 |
coding
| Benchmark | GPT-4o | Kimi K2 Instruct |
|---|---|---|
| SWE-bench Verified | 33.2 | — |
| TerminalBench | — | 30 |
multimodal
| Benchmark | GPT-4o | Kimi K2 Instruct |
|---|---|---|
| MMMU | 72.2 | — |
| MMMU-Pro | 59.9 | — |
| CharXiv Reasoning | 58.8 | — |
agent
| Benchmark | GPT-4o | Kimi K2 Instruct |
|---|---|---|
| TAU-Bench Retail | 60.3 | — |
math
| Benchmark | GPT-4o | Kimi K2 Instruct |
|---|---|---|
| AIME 2025 | — | 49.5 |
Category Winners
coding
M
Kimi K2 Instruct32.5
math
M
Kimi K2 Instruct7.9
reasoning
M
Kimi K2 Instruct36.9
general
46.2
multimodal
36.0
agent
28.4