# Claude Opus 5.5 Tops Mazur Constrained Story Writing Benchmark

> Source: <https://runtimewire.com/article/claude-opus-5-5-tops-mazur-constrained-story-writing-benchmark>
> Published: 2026-10-04 20:50:52+00:00

# Claude Opus 5.5 Tops Mazur Constrained Story Writing Benchmark

**Anthropic’s Claude Opus 5.5 scored 1.00 on the 50-task sample at $0.0242 per task. OpenAI’s GPT-6 Astra and GPT-6.1 Sol Pro tied for second at 0.99, with Sol Pro costing $0.0121 per task.**

        By [Ryan Merket](https://runtimewire.com/author/ryan-merket)
        · Published 

### Leaderboard

| Rank | Model | Mean score | Est. cost/task | 
|---|---|---|---|
| 1 | Anthropic: [Claude Opus 5.5](https://runtimewire.com/models/anthropic/claude-opus-5.5) | 1.00 | $0.0242 | 
| 2 | OpenAI: [GPT-6 Astra](https://runtimewire.com/models/native-openai/gpt-6-astra-576a599a64cb3dde) | 0.99 | $0.0599 | 
| 3 | OpenAI: [GPT-6.1 Sol](https://runtimewire.com/models/native-openai/gpt-6.1-sol-dedfefd51120f283) Pro | 0.99 | $0.0121 | 
| 4 | Anthropic: Claude Sonnet 5.5 | 0.98 | $0.0106 | 
| 5 | Z.ai: GLM 5.3 Prime | 0.97 | $0.0111 | 
| 6 | xAI: Grok Latest | 0.84 | $0.0077 | 
| 7 | Google: Gemini 3.8 Flash | 0.70 | $0.0047 | 
| 8 | Qwen: Qwen3.8 Max Prime | 0.66 | $0.0141 | 
| 9 | StepFun: Step 3.7 Flash | 0.49 | $0.0014 | 

### How we scored it

Every model answered the same 50-task battery from **Mazur Constrained Story Writing (50-task sample)** (Lech Mazur — LLM Creative Story-Writing Benchmark; published constrained briefs and archived V4 grading protocol), one task at a time, with no tools and no retries on content.

50 open-ended tasks were graded 0–10 by OpenAI: GPT-5.6 Sol Pro against a fixed rubric — blind to which model wrote the answer — and normalized to the same 0–1 scale.

Reasoning effort was pinned to **max** for every model whose lane exposes a control (OpenAI-style `reasoning_effort`, Anthropic extended thinking, Gemini thinking config). Models marked *vendor default* on the interactive board expose no such control and ran as shipped.

A model's **mean score** averages its graded tasks; a generation failure counts as 0. **Cost per task** is estimated from each model's published per-token pricing ("—" where pricing isn't public), so treat it as directional, not billing-exact.

Grading could not be completed for every task on 9 models; their means cover the graded subset.

Explore every prompt, answer, and per-task grade in the [interactive leaderboard](https://runtimewire.com/benchmarks/mazur-writing-leaderboard).