Claude has long been the model writers reach for, and Opus 5 continues that — clearer, warmer, and less hedged than the models before it, with fewer of the tics that make AI prose recognisable. The catch is length: left unbriefed, it writes more than you asked for.
This page is written for writers, editors, and content teams. If you want the model overview first, start with the complete Claude Opus 5 guide.
| Best effort level | medium |
| Context available | 1M tokens — enough for real working material, not just snippets |
| Cost | $5 per million input tokens, $25 per million output |
Why Claude Opus 5 suits long-form writing
- Prose quality is high, with noticeably fewer hedging phrases and filler transitions.
- It holds a voice across a long piece when you give it samples to match.
- It handles structural editing — reordering an argument, cutting a section — not just line editing.
- It pushes back on a weak argument rather than politely improving the wording of it.
A workflow that works
- Give it three paragraphs of your own writing as a voice sample. This does more than any adjective-laden style brief.
- Specify length in words, not adjectives. "Concise" is interpreted; "600 words" is obeyed.
- Ask for the outline first on anything over a thousand words, and edit the outline rather than the draft.
- Use it as an editor, not just a writer. Ask what to cut — its answer is usually right and usually painful.
- Write the opening yourself if voice matters most there; hand it the rest.
Settings to start from
Set the effort level to medium. Writing rarely needs the top of the effort ladder. medium is fast and strong; step up to high for technically demanding pieces where accuracy carries the argument.
Leave adaptive thinking on — it is the default on Claude Opus 5, and turning it off introduces failure modes that are not worth the saving. See effort levels explained if you want to tune this properly.
What to watch out for
- Default length runs long. An explicit conciseness instruction typically cuts output length by around a fifth; a word count cuts it further.
- Written deliverables run longer still. Files it writes to disk — reports, briefs, documentation — tend toward padding sections unless you cap them.
- It is not a fact-checker. Verify anything specific before publishing.
The verdict
The best drafting-and-editing partner in the Claude line — as long as you tell it how long the piece should be.
If cost is the constraint, read how to cut Claude Opus 5 costs before downgrading — routing and caching usually beat switching models.
More on Claude Opus 5
Start with the complete Claude Opus 5 guide for the overview, or go deeper:
Ready to go deeper?
Read the full Claude Opus 5 guideFrequently Asked Questions
Why does Claude Opus 5 write such long responses?
Longer default output is a deliberate characteristic of the model, and lowering the effort level does not reliably shorten it. Prompting does: a short conciseness instruction cuts user-facing length meaningfully, and an explicit word count works better still.
Can Claude Opus 5 match my writing voice?
Reasonably well, and better from examples than from description. Three paragraphs of your own writing as a sample outperforms any list of style adjectives. For a recurring voice, keep the sample in a reusable system prompt.
Is Opus 5 worth it for writing, or is a cheaper model enough?
For routine content, Sonnet 5 at $3/$15 is often indistinguishable. Opus 5 earns its price on pieces where the argument is the product — technical explainers, strategy documents, anything where being wrong is worse than being late.
