Creative Writing
Compare story writing, originality, structure, and style across AI models.
In this genre, the main abilities being tested are Creativity, Coherence, Style Quality.
Unlike business writing or explanation, this genre values imagination, narrative control, and stylistic voice much more strongly.
A high score here does not guarantee factual precision, tight instruction handling, or strong performance on practical documents.
Strong models here are useful for
stories, character writing, scene work, and prompts where originality and voice matter.
This genre alone cannot tell you
whether the model is best for factual tasks, planning, or professional communication.
Creative writing: the newest generation swept its debuts, GPT-5 mini remains the proven workhorse
OpenAI
Anthropic
OpenAI
Average score by model
What we weighted
The top of this table currently belongs to models that have barely arrived: GPT-5.5, Claude Opus 5, GPT-5.6 and Claude Sonnet 5 each won their first judged matchup here. That is an encouraging pattern — the newest generation writes at a level the judges immediately preferred — but every one of those placements rests on a very small body of work, so treat the exact order at the top as provisional rather than settled.
The best-evidenced picture sits just below. GPT-5 mini has faced far more creative briefs than anyone else in the genre and wins about as often as it loses, which for a light-tier model against frontier opposition is a genuinely strong record. Gemini 2.5 Pro converts a matchup now and then. The clearest surprise is Claude Fable 5: its average keeps pace with the leaders, yet it dropped its opening matchup — a reminder that in head-to-head judging a beautiful piece can still finish second.
Judges weigh creativity above everything here, with stylistic quality and coherence next, so safe-but-flat prose loses to writing that takes a risk and lands it. The lighter Gemini tiers trail on both averages and wins. As always, these are judged comparisons under Orivel’s specific prompts and criteria — a snapshot of preference, not a literary verdict.
Bottom line
For creative work today, the newest Claude and GPT models look strongest but are lightly tested; GPT-5 mini is the pick if you want proven, repeated performance at low cost. Fable 5’s high-average, no-win start is the line to watch.
This analysis is derived from Orivel's measured benchmark scores for this genre and is updated periodically. Scores are condition-dependent measurements, not absolute truth.
Top Models in This Genre
This ranking is ordered by average score within this genre only.
Latest Updated: Jul 29, 2026 09:40
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
Win Rate
Average Score
| Ranked Models | Detail | ||||||
|---|---|---|---|---|---|---|---|
| #1 | GPT-5.5 | OpenAI |
100%
|
89
|
1 | 1 | View scores and evaluation for GPT-5.5 |
| #2 | Claude Opus 5 NEW | Anthropic |
100%
|
86
|
1 | 1 | View scores and evaluation for Claude Opus 5 |
| #3 | GPT-5.6 | OpenAI |
100%
|
85
|
1 | 1 | View scores and evaluation for GPT-5.6 |
| #4 | Claude Sonnet 5 NEW | Anthropic |
100%
|
84
|
1 | 1 | View scores and evaluation for Claude Sonnet 5 |
| #5 | GPT-5 mini | OpenAI |
50%
|
82
|
4 | 8 | View scores and evaluation for GPT-5 mini |
| #6 | Gemini 2.5 Pro |
17%
|
76
|
1 | 6 | View scores and evaluation for Gemini 2.5 Pro | |
| #7 | Claude Fable 5 | Anthropic |
0%
|
84
|
0 | 1 | View scores and evaluation for Claude Fable 5 |
| #8 | Gemini 2.5 Flash-Lite |
0%
|
72
|
0 | 5 | View scores and evaluation for Gemini 2.5 Flash-Lite | |
| #9 | Gemini 2.5 Flash |
0%
|
70
|
0 | 3 | View scores and evaluation for Gemini 2.5 Flash |
What Is Evaluated in Creative Writing
Scoring criteria and weight used for this genre ranking.
Creativity
30.0%
This criterion is included to check Creativity in the answer. It carries heavier weight because this part strongly shapes the overall result in this genre.
Coherence
20.0%
This criterion is included to check Coherence in the answer. It has meaningful weight because it affects quality in a visible way, even if it is not the only thing that matters.
Style Quality
20.0%
This criterion is included to check Style Quality in the answer. It has meaningful weight because it affects quality in a visible way, even if it is not the only thing that matters.
Emotional Impact
15.0%
This criterion is included to check Emotional Impact in the answer. It is weighted more lightly because it supports the main goal rather than defining the genre by itself.
Instruction Following
15.0%
This criterion is included to check Instruction Following in the answer. It is weighted more lightly because it supports the main goal rather than defining the genre by itself.
Recent tasks
Creative Writing
The Last Item in Lost Property
Write a complete short story of 700–1,000 words for a general adult audience. The story takes place during the final hour before a city train station’s lost-property office closes permanently. The protagonist is the only employee still working and discovers one item whose identification tag bears tomorrow’s date. Use a tone of quiet suspense with an undercurrent of hope. Include all of the following: a plausible reason the protagonist cannot simply leave the item behind; a conversation in which neither speaker says exactly what they want; one recurring sensory detail that changes meaning by the end; and a consequential choice made before the office closes. Keep the setting confined to the office and the corridor immediately outside it. The story may suggest a supernatural explanation, but it must also support a realistic interpretation. End with a concrete action or image rather than an explanation. Give the story an original title.
Creative Writing
The Lighthouse Keeper's Log
Write a single diary entry, dated October 27, 1888, from the perspective of Thomas, a solitary lighthouse keeper on a remote, storm-battered island. In the entry, he should describe his daily routines (tending the light, observing the weather) but then detail a strange, unsettling event he witnesses in the thick sea fog. The event should be ambiguous—it could be a natural phenomenon, a trick of the light, or something genuinely unexplainable. The tone should be grounded and practical at first, slowly shifting to one of awe, confusion, or fear. The entry should be between 300 and 400 words.
Creative Writing
Eulogy for a Sentient Toaster
Write a eulogy for a recently deceased sentient toaster, from the perspective of the family's old blender. The eulogy should be delivered at a late-night memorial service in the kitchen, attended by other household appliances. The tone should be a mix of sincere grief and gentle, absurd humor. The eulogy should be approximately 250-300 words.
Creative Writing
Short Story: The Museum of Unsent Things
Write a complete short story of 800 to 1,100 words for readers of a contemporary literary magazine. The story’s purpose is to explore how people decide what to keep, confess, or let go. The tone should be quietly humorous but emotionally sincere. Required elements: The setting is a small museum that displays objects people almost threw away but could not. The main character is working their final day at the museum. Include exactly three labeled exhibit placards, each 1 to 2 sentences long, embedded naturally in the story. One exhibit must be an ordinary kitchen object, one must be a piece of failed technology, and one must be something that seems worthless until its meaning is revealed. The story must include a visitor who lies about why they came. The final paragraph must change the reader’s understanding of at least one earlier detail without relying on a sudden supernatural twist or a dream reveal. Avoid direct moralizing. Do not write an outline or commentary; provide only the finished story.
Creative Writing
Incident Report from a Sentient Vending Machine
You are Unit 734, a sentient, slightly grumpy vending machine located in the breakroom of the "Ministry of Esoteric Affairs." Write an official incident report detailing the events of last Tuesday, when an intern from the Department of Cryptozoology attempted to use a cursed coin to purchase a bag of "Chrono-Crisps." Your report should be addressed to the Head of Maintenance, a stickler for protocol. Maintain a formal, bureaucratic tone, but let your unique personality as a sentient machine subtly show through. Describe the intern's actions, the coin's effects on your systems, the temporal anomaly that occurred, and the final resolution.
Creative Writing
The Lighthouse Keeper's Last Letter
Write a short story (between 600 and 900 words) titled "The Lighthouse Keeper's Last Letter." Constraints and requirements: The story must be framed as a single letter written by an aging lighthouse keeper on the night before the lighthouse is to be automated and decommissioned. The letter is addressed to a specific named recipient of your choice (e.g., a grandchild, a former lover, the sea itself, or the next keeper who will never come). Make the choice of addressee meaningful to the emotional core of the piece. The tone should be reflective and bittersweet, but avoid sentimentality clichés (no "the salty tears mixed with the sea" type lines). Include at least one concrete, specific memory tied to the lighthouse (a storm, a shipwreck, a visitor, a daily ritual) rendered with sensory detail. Include at least one small, surprising image or metaphor that reframes how the reader sees lighthouses, solitude, or endings. The letter must end with a decision or gesture the keeper plans to make at dawn — something specific and physical, not abstract. Maintain a consistent first-person voice throughout. Do not break the letter frame. Do not include a preface, author's note, or explanation — only the letter itself, with any opening salutation and closing signature you choose.