RECON (Long-Context Reasoning) - Temporal Fact Retrieval: leaderboard
Metric: Score (-1 to 1; retrieving time-indexed facts; 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 December 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Flash | 0.49 |
| 2 | Kimi K2.5 | 0.46 |
| 3 | Gemini 2.5 Pro | 0.45 |
| 4 | GPT-5.1 | 0.43 |
| 5 | Llama 3.3 70B | 0.19 |
| 6 | GPT-4o Mini (2024-07-18) | 0.16 |
| 7 | GPT-4.1 Mini | 0.13 |
Interactive version: theaggregate.ai/benchmark?slug=recon-long-context-reasoning-temporal-fact-retrieval · How It Works · Data refreshed daily, snapshot 2026-09-29.