TIDE (Evolving Documents, Closed-Book): leaderboard

Metric: Macro-averaged strict accuracy (%; mean over the eight task types; TIDE, 3,050 expert-verified questions over 644 Bangladesh customs instruments (1969-2025); free-form answers are judged by a council of Claude Opus 4.6, Gemini 3.5 Flash and GPT-5.1 by majority vote with a hard date gate, deterministic tasks by exact match; strict accuracy without the partial-rubric credit; thinking enabled, temperature 0; closed book: no document is supplied). Source: arxiv.org. Saturation forecast: Around 2030. 9 models tracked.

Top models

#ModelScore
1GPT-5.2 (Thinking)35.39
2Gemini 3.5 Flash34.15
3Kimi K2.5 (Thinking)33.28
4Gemma 4 31B (Reasoning)32.04
5GPT-OSS-120B30.76
6Qwen 3.6 35B A3B26.39
7Claude Sonnet 4.5 (Thinking)25.91
8GLM-4.625.54

Interactive version: theaggregate.ai/benchmark?slug=tide-evolving-documents-closed-book · How It Works · Data refreshed daily, snapshot 2026-09-29.