LongNovel (16K): leaderboard
Metric: Balanced accuracy (%; S, novels of about 16k tokens context scale; binary detection of hallucinated summaries of Chinese novels and English BookSum chapters (8 hallucination types, human-revised test set); direct prompt with the summary at the end; an output without a valid consistency score counts as an incorrect prediction; chance is 50). Source: arxiv.org. Saturation forecast: Around December 2026. 12 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek V4 Flash | 78.5 |
| 2 | GPT-5.2 Instant | 78.5 |
| 3 | Claude Sonnet 4.5 | 75.5 |
| 4 | Qwen 3 32B | 59 |
| 5 | Qwen 3 14B | 58.8 |
| 6 | Qwen 3 8B | 56.8 |
| 7 | internlm2.5-20B-chat | 54.5 |
| 8 | Llama 3.1 8B Instruct | 52.5 |
| 9 | GLM-4 9B Chat | 50.5 |
Interactive version: theaggregate.ai/benchmark?slug=longnovel-16k · How It Works · Data refreshed daily, snapshot 2026-09-29.