Khondo (MonoSeq) - Page Typing: leaderboard

Metric: Page accuracy (%; share of pages assigned their correct administrative domain; forms of one domain concatenated in order; 210 test packets of 5 to 20 page images of Bangladeshi government forms (Bangla and English, 14 domains); zero-shot order-aware prompt, all page images in one request, JSON output, temperature 1.0, reasoning disabled, invalid outputs retried up to three times). Source: arxiv.org. Saturation forecast: Around December 2026. 5 models tracked.

Top models

#ModelScore
1Qwen 3.6 Plus (Non-reasoning)57.9
2GPT-5.4 (Non-reasoning)52.9
3Kimi K2.5 (Non-reasoning)51.6
4Gemini 3.5 Flash (Minimal)50.1
5GLM-4.6V (Non-reasoning)31.8

Interactive version: theaggregate.ai/benchmark?slug=khondo-monoseq-page-typing · How It Works · Data refreshed daily, snapshot 2026-09-29.