# RealWorldQA → AI2D_TEST

Calibration · https://benchgap.net/calibration/realworldqa/current/to/ai2dtest/current · data retrieved 2026-11-18

AI2D_TEST is estimated from RealWorldQA with a offset logistic curve fitted on 7 models measured on both: y = 0.8760 + (0.9325 − 0.8760) / (1 + exp(−200.00·(x − 0.8087))), R² = 0.94, cross-validated error 1.4 pp. It is used for 4 estimates.

| Estimated model | RealWorldQA | AI2D_TEST | Source |
|---|---|---|---|
| [Ternary Bonsai 2 27B](https://benchgap.net/model/ternary-bonsai-2-27b.md) | 80.1% | 88.7% | estimated ± 1.4 pp, medium confidence |
| [LFM2.5-VL-3B](https://benchgap.net/model/lfm2-5-vl-3b.md) | 73.1% | 87.6% | estimated ± 1.4 pp, medium confidence |
| [LFM2.5-VL-450M](https://benchgap.net/model/lfm2-5-vl-450m.md) | 58.4% | 87.6% | estimated ± 1.4 pp, low confidence |
| [North Micro Vision Instruct](https://benchgap.net/model/north-micro-vision-instruct.md) | 62.2% | 87.6% | estimated ± 1.4 pp, low confidence |

## Fitted on

| Model | RealWorldQA | AI2D_TEST |
|---|---|---|
| [Claude Opus 4.5](https://benchgap.net/model/claude-opus-4-5.md) | 77.0% | 87.7% |
| [Gemini 3 Pro](https://benchgap.net/model/gemini-3-pro.md) | 83.3% | 94.1% |
| [GPT-5.2](https://benchgap.net/model/gpt-5-2.md) | 83.3% | 92.2% |
| [Kimi K2.5](https://benchgap.net/model/kimi-k2-5.md) | 81.0% | 90.8% |
| [Qwen3.5 397B](https://benchgap.net/model/qwen3-5-397b.md) | 83.9% | 93.9% |
| [Qwen3.6-35B-A3B](https://benchgap.net/model/qwen3-6-35b-a3b.md) | 85.3% | 92.7% |
| [ZAYA1-VL-8B](https://benchgap.net/model/zaya1-vl-8b.md) | 65.0% | 87.5% |
