Fiction Eval leaderboard · genre
Best AI model for cozy mystery writing
July 2026 — scores recomputed using only the cozy mystery scenes (84 blind matchups, three tries per model). small sample — treat as rough
GLM 5.2 leads, and comes out ahead of #2 in 68% of re-runs. This cozy mystery-only sample is small, so treat the exact order as rough and check where the ranges overlap before reading much into close neighbors.
| # | Model | Cozy Mystery score (likely range) |
|---|---|---|
| 1 | GLM 5.2 | 1710 (likely 1659–1752) |
| 2 | Claude Sonnet 4.6 | 1694 (likely 1645–1738) |
| 3 | Claude Opus 4.8 | 1664 (likely 1610–1706) |
| 4 | Grok 4.3 | 1620 (likely 1548–1687) |
| 5 | Gemini 3.1 Pro | 1494 (likely 1421–1564) |
| 6 | DeepSeek V3.2 | 1329 (likely 1262–1404) |
| 7 | Mistral Large | 1329 (likely 1259–1404) |
| 8 | Llama 4 Maverick | 1159 (likely 1106–1245) |
GLM 5.2 on the cozy cold open scene
I was arranging the window display at Tides & Tomes—three nautical mysteries, a lighthouse cookbook, and a ceramic seal I'd been trying to sell since the Clinton administration—when I heard the scream. Not a blood-curdling scream. More of a surprised-yoghurt scream, the kind you make when life does something unexpected and you haven't had your second cup of coffee. This was Delia, who ran the flower stall on the harbor side, and she had no talent whatsoever for theater, so I knew it was real. I found her standing in the alley between my shop and Sugar & Salt, backing into a stack of empty crates. "It's Margie," she said. "In the freezer." I should explain that Margie Dupont was not the sort of woman you found dead. She was the sort of woman who found *you* dead, revived you with espresso, and then lectured you about your cholesterol. She had run Sugar & Salt for thirty-one years, and the town had collectively gained an estimated four hundred pounds under …
Read every model's take side by side on the board.
More
The full 8-model board · the July 2026 report · all head-to-heads
How it works: every pair of models is judged blind on the same scene, with the passages'
order flipped so being shown first can't sway it, by GPT-5.4 — a family that isn't on the
board, so nobody scores their own side. Each score carries a likely range; overlapping ranges
are called a tie. Slop is scored separately by a fixed checklist, not an AI. Full board,
every prompt, and the FAQ: thebookfactoryai.com/board.
Get each new board by email on the model-drop list.
Writing a book of your own? Book Factory runs the same craft checks on full manuscripts.
Writing a book of your own? Book Factory runs the same craft checks on full manuscripts.