PredicateLongBench - Lexicographic Longest (Synthetic): leaderboard

Metric: Accuracy (%; share of the 100 synthetic word lists of about 128K tokens whose answer is exactly the one word sequence satisfying the predicate, retrieve the longest lexicographically sorted run; up to 16K output tokens including reasoning, an answer past the cap counts as wrong). Source: arxiv.org. Saturation forecast: Around September 2028. 9 models tracked.

Top models

#ModelScore
1Claude Opus 4.6 (Thinking)23
2Gemini 3.1 Pro (Preview) (Low)15
3GPT-5.4 (High)12
4Claude Opus 4.65
5MiniMax-M2.72
6Gemini 3.1 Pro (Preview) (High)2
7Qwen 3.5 397B A17B0
8GLM-5.10
9GPT-5.4 (Non-reasoning)0

Interactive version: theaggregate.ai/benchmark?slug=predicatelongbench-lexicographic-longest-synthetic · How It Works · Data refreshed daily, snapshot 2026-09-29.