The catch, stated plainly
What you actually pay for
You pay the model, and you pay it directly. We never touch the money and we take no cut of it.
Writing a chapter means asking a large language model to write it, and those models charge per token. You bring your own OpenRouter key, top it up with however much you want, and the bill goes from them to you.
That arrangement is deliberate: the one place you should have real freedom is the money. Choose a model that costs a tenth of a cent a chapter or one that costs ten cents. Change it whenever. Nothing here marks it up, and nothing here decides for you.
Then we work to make that bill smaller. The app pins providers so prompt caching actually lands, compacts old chapters into summaries so a 200-chapter novel does not re-read itself, and shows you the running cost per turn while you write.
What will your novel cost?
Move the sliders. The maths below replays the exact sequence of calls the app makes for each chapter — the first draft, your revisions, the summary written on accept, and any searches through earlier chapters — then prices every one of them at the provider's real rates, caching included.
Your novel
Live prices from OpenRouterOrdered by what fiction writers on OpenRouter actually use.
Estimated total
$0.776
$0.016 per chapter · paid to OpenRouter, not to us
Priced on CoreWeave, the provider we route this model to.
Caching
Reuses 91% of prompt tokens
Saves $0.547 — without caching this novel would cost $1.32.
Where it goes
- Writing chapters
- $0.208
- Revisions
- $0.454
- Chapter summaries
- $0.023
- Searching earlier chapters
- $0.092
- Tokens in / out
- 8.6M / 424.1k
Cost per chapter
An estimate, not a quote. It replays the real call sequence at this provider’s published rates; actual tokenisation and output length vary per model.
Best bang for the buck
Three good defaults, priced on a 50-chapter novel at 2,000 words a chapter with two revisions each.
DeepSeek: DeepSeek V4 Flash 0423
$0.776
for the whole novel · $0.016 a chapter
Fast, genuinely cheap, and it caches properly. What most people should start on — a full novel costs less than a coffee.
Reuses 91% of prompt tokens from cache.
Z.ai: GLM 5.2
$2.72
for the whole novel · $0.054 a chapter
Noticeably stronger prose and continuity for a few times the price, and it still reuses most of the history between chapters.
Reuses 91% of prompt tokens from cache.
SpaceXAI: Grok 4.5
$19.76
for the whole novel · $0.395 a chapter
The strongest writing on the list — but xAI does not reuse a growing prompt, so a long novel re-reads itself at full price every chapter.
No prefix reuse — the price above already reflects that.
Prices from OpenRouter, on the provider we route to. You are never locked to these — all 400+ models are one dropdown away.