Eye-Q - Iterative Refinement - Persian: leaderboard

Metric: Exact-Match Accuracy (%; up to three attempts, told when wrong, Persian subset). Source: arxiv.org. Saturation forecast: Around December 2026. 6 models tracked.

Top models

#ModelScore
1Gemini 2.5 Pro41.23
2Grok 4.1 Fast (Reasoning)24.29
3Gemini 2.5 Flash17.84
4GPT-5.2 (Non-reasoning)11.09
5Qwen 3 VL 235B A22B Instruct4.2
6Llama 4 Scout3.77

Interactive version: theaggregate.ai/benchmark?slug=eye-q-iterative-refinement-persian · How It Works · Data refreshed daily, snapshot 2026-09-25.