Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Claude Haiku 4.5 | Qwen3-235B-A22B-Thinking-2507 |
|---|---|---|
| GPQA Diamond | 73 | 81.1 |
| Humanity's Last Exam | — | 18.2 |
math
| Benchmark | Claude Haiku 4.5 | Qwen3-235B-A22B-Thinking-2507 |
|---|---|---|
| AIME 2025 | 80.7 | 92.3 |
coding
| Benchmark | Claude Haiku 4.5 | Qwen3-235B-A22B-Thinking-2507 |
|---|---|---|
| SWE-bench Verified | 73.3 | — |
| TerminalBench | 41 | — |
general
| Benchmark | Claude Haiku 4.5 | Qwen3-235B-A22B-Thinking-2507 |
|---|---|---|
| MMMLU | 83 | — |
agent
| Benchmark | Claude Haiku 4.5 | Qwen3-235B-A22B-Thinking-2507 |
|---|---|---|
| OSWorld | 50.7 | — |
| TAU-Bench Retail | — | 67.8 |
Category Winners
coding
70.2
math
Q
Qwen3-235B-A22B-Thinking-250774.0
reasoning
59.1
general
48.8
agent
57.3