RECON (Long-Context Reasoning): leaderboard

Metric: Score (-1 to 1; mean per-question score over all six tasks; full 50K-100K-token case file in context; 1,414 questions after a closed-book contamination filter; abstention allowed, MCQ-single wrong answers -0.2, MCQ-multiple option-weighted in [-1, 1], ordering by Kendall tau, free-form by two cross-family LLM judges; temperature 0). Source: arxiv.org. Saturation forecast: Around 2029. 8 models tracked.

Top models

#ModelScore
1GPT-5.10.29
2Gemini 2.5 Pro0.27
3Kimi K2.50.26
4Gemini 2.5 Flash0.26
5GPT-4o Mini (2024-07-18)0.23
6GPT-4.1 Mini0.21
7Llama 3.3 70B0.19

Interactive version: theaggregate.ai/benchmark?slug=recon-long-context-reasoning · How It Works · Data refreshed daily, snapshot 2026-09-29.