YZ Index
Evaluation Data
Currently showing:Run #397 | 2026-10-26 | 128 题库 | Formula v7 | Judge set v6.5
Data Disclosure:To prevent benchmark contamination and overfitting, question texts and expected answers are not disclosed. This page shows model responses, scores, and judging methods for transparency. For the full methodology, seeMethodology page。
| Model | Code Execution | Grounding | Engineering Judgment | Task Communication | Integrity | Overallpts | Value | Stability | Availability | Per Task |
|---|---|---|---|---|---|---|---|---|---|---|
| GPT-6.1 Sol gpt | 93.00 | 68.20 | 89.60 | 86.30 | 86.70 pass | 81.84 | 39.4 | 44.4 | 100.0 | |
| Claude Opus 4.7 claude | 76.90 | 83.30 | 78.10 | 58.30 | 90.00 pass | 79.78 | 5.3 | 32.8 | 100.0 | |
| Doubao Pro doubao | 87.40 | 63.10 | 87.80 | 84.00 | 86.70 pass | 76.47 | 95.7 | 42.5 | 97.7 | |
| GPT-6 Luna gpt | 84.80 | 66.00 | 83.20 | 83.80 | 78.30 pass | 76.34 | 100.0 | 38.0 | 99.0 | |
| GPT-6 Astra gpt | 84.40 | 63.90 | 90.00 | 86.10 | 86.70 pass | 75.18 | 7.8 | 35.9 | 100.0 | |
| GPT-6 Sol gpt | 85.60 | 60.20 | 87.50 | 88.80 | 86.70 pass | 74.17 | 36.9 | 36.4 | 100.0 | |
| Claude Sonnet 4.6 claude | 76.10 | 67.40 | 70.30 | 76.90 | 86.70 pass | 72.19 | 23.9 | 31.5 | 100.0 | |
| GPT-5.5 gpt | 80.10 | 60.50 | 90.00 | 90.80 | 81.70 pass | 71.28 | 17.9 | 33.9 | 97.0 | |
| GPT-o3 gpt | 65.00 | 71.90 | 84.60 | 89.00 | 78.30 pass | 68.11 | 8.6 | 31.5 | 96.0 | |
| Grok 4 grok | 69.80 | 60.80 | 52.10 | 66.50 | 78.30 pass | 65.75 | 21.3 | 24.6 | 100.0 | |
| DeepSeek V4 Pro deepseek | 62.10 | 68.50 | 81.70 | 78.20 | 81.70 pass | 64.98 | 43.9 | 40.4 | 87.4 | |
| Gemini 3.1 Pro gemini | 52.40 | 69.40 | 82.80 | 63.30 | 83.30 pass | 60.05 | 21.9 | 16.8 | 100.0 | |
| Gemini 2.5 Pro gemini | 52.60 | 62.50 | 69.30 | 61.50 | 93.30 pass | 57.06 | 31.6 | 31.3 | 87.0 | |
| GLM-4.6 zhipu | 31.80 | 69.40 | 20.00 | 20.80 | 50.00 warn | 48.72 | 25.1 | 33.8 | 54.0 | |
| Qwen3 Max qwen | - | - | - | - | - | - | 0.0 | 75.0 | 0.0 |
API Access:For programmatic access to evaluation data, please use our API。