Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| GPQA Diamond | 73 | 75.4 |
math
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| AIME 2025 | 80.7 | 70.5 |
coding
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| SWE-bench Verified | 73.3 | 72.7 |
| TerminalBench | 41 | 35.5 |
general
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| MMMLU | 83 | 86.5 |
agent
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| OSWorld | 50.7 | — |
| TAU-Bench Retail | — | 80.5 |
multimodal
| Benchmark | Claude Haiku 4.5 | Claude Sonnet 4 |
|---|---|---|
| MMMU | — | 74.4 |
Category Winners
coding
70.5
math
53.0
reasoning
66.9
general
62.9
multimodal
68.6
agent
83.4