Fiction Eval leaderboard · comparison

Grok 4.3 vs Llama 4 Maverick for fiction writing

Judged blind on fresh genre scenes (LitRPG, cozy mystery, romance, thriller, xianxia…), July 2026 — three tries per model per scene, passages shown in both orders so position can't sway it, graded by a third AI that isn't either of them.

Head-to-head

Grok 4.3 won the direct matchup 30–0 (with 0 ties) across every prompt and sample.

#ModelScoreRecordSlop
3Grok 4.31856 (likely 1815–1899)137–28–429.43
8Llama 4 Maverick686 (likely 627–741)0–206–122.92

The score is each model's overall strength across the whole 8-model field (like a chess rating); the range next to it is where the true score most likely sits. Slop is a separate count of AI tells (lower is cleaner) — it isn't part of the ranking.

The same scene, both models

LitRPG — opening of the "party aftermath" scene. Full passages for all models are on the board.

Grok 4.3:

The last health bar winked out in the corner of my vision, a red slash across the party roster that left only my name pulsing green. I knelt in the ash where the floor boss’s final nova had scorched the stone, staff still warm in my hands, every restore and barrier I had cast already logged and dismissed by the system as insufficient. [Party Wipe recorded.] [Experience penalties applied to surviving members: –40 % to next level.] [Equipment durability reduced: 12 %.] I barely registered the lines scrolling across the lower left of my sight. My fingers twitched through the empty air where the resurrection window should have appeared, but the hardcore flag had already overwritten it with something colder. [Korven: deceased. Items dropped at coordinates 47.3, 19.8.] [Mira: deceased. Items dropped at coordinates 47.3, 19.8.] [Selene: deceased. Items dropped at coordinates 47.3, 19.8.] Then the final line settled, heavier than the rest, the text itself thicker, outlined in the dull iron color the system used only for irreversible losses. …

Llama 4 Maverick:

I slumped against the cold stone wall, my back aching from the impact. The party's final moments replayed in my mind like a gruesome highlight reel: Erebus's desperate attempt to tank the boss's enraged cleave, Lyra's frantic healing attempts as our DPS struggled to keep up, and Kael's... Kael's final, doomed attempt to solo the add that had spawned during the fight. The memory of it all made my stomach churn. As the darkness began to recede, the System's notifications flooded my vision. **Combat Log: Party Wipe** * Erebus (Tank): Killed by Voidreaver's Cleave (Critical Hit) * Lyra (Healer): Killed by Voidreaver's Shadowbolt (Critical Hit) * Kael (DPS): Killed by Adds: Voidspawn (x3) * Zen (DPS): Killed by Voidreaver's Devouring Dark (Instant Kill) * Mira (Support Healer): Survived **Penalty Assessment: Party Wipe** * 10% Penalty to EXP for 24 hours * 5% Penalty to Reputation with Voidreaver's Lair for 7 days The messages continued to scroll, but my gaze locked onto the party roster. **Party Roster: Offline** * Erebus: Character …

More

Grok 4.3 full review · Llama 4 Maverick full review · the July 2026 report · all comparisons

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.