ExpertLongBench (Checklist F1): leaderboard

ExpertLongBench v1 (arXiv:2506.01241v3): mean CLEAR structured-checklist F1 score (0-100%) across 11 expert-level long-form generation tasks spanning 9 domains (outputs up to >5,000 tokens, greedy decoding temp=0); higher is better.

Metric: Mean Checklist F1 (%). Source: huggingface.co. Saturation forecast: Around 2031. 13 models tracked. Updated 21 Nov 2026.

Top models

#ModelScoreOverall rank
1Gemini 2.5 Pro33.4#276
2GPT-531#178
3Gemini 2.0 Flash26.8#723
4GPT-4o (2024-11-20)26.5#859
5GPT-4o Mini (2024-07-18)26.1#1484
6Llama 3.3 70B Instruct24.6#1385
7Mistral Large 2 (Nov) Instruct (2411)24#1217
8Qwen 2.5 72B23.8#1188
9Mistral Nemo Instruct (2407)23.8#3118
10Llama 3.1 8B Instruct23.4#4600
11Claude 3.7 Sonnet23.2#470
12Qwen 2.5 7B20.5#2929
13Claude 3.5 Haiku19.3#1580

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

Interactive version: theaggregate.ai/benchmark?slug=expertlongbench-checklist-f1 · How It Works · Data refreshed daily, snapshot 2026-11-21.