AA Long Context Reasoning — leaderboard
AA-LCR: ability to extract, reason about, and synthesize information from documents of 10k-100k tokens (100 questions). Proprietary AA benchmark.
Metric: Accuracy (%). Source: artificialanalysis.ai. Status: saturation imminent. 472 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.2 Codex (xHigh) | 75.67 |
| 2 | GPT-5 (High) | 75.6 |
| 3 | GPT-5.1 (High) | 75 |
| 4 | Kimi K3 | 74.67 |
| 5 | GPT-5.5 (xHigh) | 74.33 |
| 6 | MiniMax-M3 | 74 |
| 7 | GPT-5.4 (xHigh) | 74 |
| 8 | Claude Opus 4.5 (Thinking) | 74 |
| 9 | KAT-Coder-Pro V1 | 74 |
| 10 | GPT-5.6 Terra (Max) | 74 |
| 11 | GPT-5.6 Luna (Max) | 74 |
| 12 | GPT-5.3 Codex (xHigh) | 74 |
| 13 | GPT-5.6 Sol (Max) | 73.67 |
| 14 | MiMo-V2.5-Pro | 73.33 |
| 15 | GPT-5.5 (High) | 73.33 |
Interactive version: theaggregate.ai/benchmark?slug=aa-long-context-reasoning · How the rankings work · Data refreshed daily, snapshot 2026-07-22.