Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| BigBench-Hard | 81.2 | — |
| GPQA Diamond | 46.7 | — |
| ARC-Challenge | 94.8 | — |
| Humanity's Last Exam | — | 13.6 |
general
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| IFEval | 83.6 | — |
| Chatbot Arena Elo | 1207 | — |
| MMLU-Pro | 66.4 | — |
| SimpleQA | — | 44.4 |
coding
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| HumanEval | 80.5 | — |
language
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| HellaSwag | 94.4 | — |
| WinoGrande | 85.3 | — |
safety
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| TruthfulQA | 60.1 | — |
multimodal
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| MMMU-Pro | — | 69.3 |
| CharXiv Reasoning | — | 66.1 |
| ScreenSpot Pro | — | 61.8 |
agent
| Benchmark | Llama 3.1 70B | Qwen3 VL 235B A22B Thinking |
|---|---|---|
| OSWorld | — | 38.1 |
Category Winners
coding
31.5
math
Q
Qwen3 VL 235B A22B Thinking69.3
reasoning
30.5
general
Q
Qwen3 VL 235B A22B Thinking59.5
language
70.8
multimodal
Q
Qwen3 VL 235B A22B Thinking45.1
safety
27.6
agent
Q
Qwen3 VL 235B A22B Thinking