Mercor AA-LCR Extended: leaderboard

Models answer long-context reasoning questions on Mercor’s held-out extension of AA-LCR. Pass@1 uses the extended task population; original public-benchmark scores are separate.

Metric: Pass@1 (%). Source: www.mercor.com. 51 models tracked.

Top models

#ModelSettingScoreOverall rank
1GPT-5.6 Pro SolMax79.5#27
2GLM-5.3Max76.8#73
3Gemini 4 ArgonHigh74.7#14
4Muse WonderMax74.4#31
5GPT-5.6 TerraMax74.4#68
6Claude Opus 5.5Max74.2#7
7Claude Opus 5Max73.9#23
8Claude Sonnet 5.5Max73.8#21
9Qwen 3.8 MaxxHigh73.3#64
10Claude Fable 5.1High73#17
11GPT-5.6 LunaMax72.2#130
12Claude Fable 5Max71.1#28
13DeepSeek V4.1 FlashMax70.8#69
14GPT-5.5xHigh70.5#50
15GPT-6 SolMax70.2#29

No result here: #1 Claude Fable 5.2, #2 GPT-6.2 Astra, #3 Gemini 4 Carbon.

Interactive version: theaggregate.ai/benchmark?slug=mercor-aa-lcr-extended · How It Works · Data refreshed daily, snapshot 2026-11-17.