Fiction.LiveBench - 16K tokens
Chat & Writing · 2025-02-19
Fiction.LiveBench quiz accuracy when the story is cut to about 16K tokens, so the relevant passages sit inside that much surrounding fiction. The same private, open-ended comprehension questions are asked at every length; question counts are not published.
Top models (higher is better)
| Model | Score |
|---|---|
| Opus 4.6 | 100.0 |
| Sonnet 4 | 100.0 |
| Gemini 3 Flash Preview | 100.0 |
| GPT-5.2 | 100.0 |
| GPT-5 | 97.2 |
| Gemini 3 Pro Preview | 96.6 |
| Opus 4.5 | 94.4 |
| Grok 4 | 94.4 |
| Grok 4 Fast | 94.4 |
| Sonoma Sky Alpha | 94.4 |
| Gemini 2.5 Pro | 91.7 |
| GLM-5 | 91.7 |
| o3 | 88.9 |
| Sonnet 4.5 | 86.1 |
| DeepSeek-V3.2 | 86.1 |
| Kimi K2.5 | 86.1 |
| DeepSeek-V3.2-Exp | 83.3 |
| GLM-4.7 | 83.3 |
| GPT-5.2 Pro | 83.3 |
| QwQ-32B | 83.3 |
| Gemini 2.5 Flash | 80.6 |
| O4 Mini | 77.8 |
| Qwen3.5 Plus (2026-02-15) | 77.8 |
| DeepSeek-R1-0528 | 75.0 |
| Qwen3 235B A22B Thinking 2507 | 75.0 |
| Qwen3 32B | 74.2 |
| DeepSeek-V3.1 | 72.2 |
| Opus 4 | 69.4 |
| MiniMax M1 | 69.4 |
| GPT-5 Mini | 69.4 |
| Qwen3-235B-A22B | 67.7 |
| ChatGPT-4o Latest (source-unspecified snapshot) | 66.7 |
| Kimi K2 Instruct 0905 | 66.7 |
| Qwen3.6 Plus (2026-04-02) | 66.7 |
| GPT-4.1 | 63.9 |
| GPT-4.5 | 63.9 |
| MiniMax M2.5 | 63.9 |
| Qwen3 14B | 62.5 |
| Qwen3 8B | 62.1 |
| GLM-4.5 | 58.3 |