Claude Haiku 4.5 vs Claude Sonnet 4.6 for Creative Writing

Creative Writing (fiction, storytelling, creative content) demands: high creative_problem_solving (fresh, feasible plot/scene ideas), persona_consistency (stable voice and character), long_context handling (scene-to-scene continuity across tens of thousands of tokens), and constrained_rewriting (compressing or adapting text into tight formats). Our Creative Writing composite is a 3-test average (creative_problem_solving, persona_consistency, constrained_rewriting) on our 1–5 scale. In our testing Sonnet 4.6 scores 4.333... overall because it scores 5/5 on creative_problem_solving, while Haiku scores 4/5. Both models score 5/5 on persona_consistency and 3/5 on constrained_rewriting. There is no external benchmark in the payload for this task, so our internal 3-test composite is the primary signal. Supporting telemetry: both models tie on long_context (5) and faithfulness (5), which explains why both maintain continuity and avoid source hallucination; Sonnet's superior creative_problem_solving explains its advantage in generating novel plot moves and inventive scene structure.

1. Brainstorming plot arcs and surprising twists — Sonnet 4.6 shines: creative_problem_solving 5 vs Haiku 4 in our tests, so Sonnet produced more specific, non-obvious ideas in our scenarios. 2) Maintaining character voice across a long manuscript — both models perform equally well: persona_consistency 5 for both and long_context 5, so either handles multi-chapter continuity. 3) Tight-format rewriting (e.g., compress a chapter to a 280‑char microfiction) — both score 3 on constrained_rewriting, so expect similar difficulty and manual editing. 4) High-volume or iterative editing on a budget — Haiku 4.5 is preferable: task score 4.00 with much lower token costs (input 1 / output 5) compared with Sonnet (input 3 / output 15). 5) Sensitive or risky content moderation while writing — Sonnet's safety_calibration is 5 vs Haiku's 2 in our tests (not part of the Creative Writing composite), so Sonnet is safer at refusing harmful prompts while allowing legitimate creative uses.

For Creative Writing, choose Claude Haiku 4.5 if you need a lower-cost, high-quality writer for bulk drafts, iterative edits, or budget-constrained workflows (Haiku task score 4.00; input/output costs 1/5). Choose Claude Sonnet 4.6 if you prioritize more original ideation and non-obvious plot solutions — Sonnet leads by 0.33 points on our Creative Writing composite (4.33 vs 4.00) and scores 5/5 on creative_problem_solving — and you're willing to pay higher token costs (input/output 3/15) and leverage a larger context window.