Anthropic’s Claude Opus 4.7 lands in fourth place on our AI writing leaderboard with an overall score of 6.44/10. It’s a solid writer that can handle most briefs. Senior writer Nick Banks put it through eight writing tasks across four capability areas. Here’s the verdict.
See the full AI writing leaderboardSales and customer messages
Score: 6.75/10
A confident showing. The ‘GreenBox’ food supplier email about discontinuing heritage tomatoes, eggs and pasta was warm and human. But perhaps took itself a little too seriously.
Nick: “On first glance this seems like a decent attempt to communicate some news that’s likely to be disappointing to the consumer in a way that’s warm, understanding. On second read though, the tone feels very heavy handed, almost grave – it’s like they’re informing the customer that someone’s died, not that they’re unable to include some tomatoes in a food box.”
The ‘Roam Presso’ product description (a hand-powered portable espresso maker for outdoor adventurers) started strongly and with proper swagger. It only tripped over itself with the bizarre ‘At just 340g, it’s lighter than your head torch and your dignity combined.’ What?
Nick’s verdict: “Lots of attitude, energy, cultural references without being too cliched or cringey. Not sure about ‘engineered for backcountry abuse’ either, but I guess it fits with the schtick. Otherwise it’s well written, punchy, articulate.”
Snappy headlines and titles
Score: 5/10
The weakest area, and a surprise given the brief practically wrote itself. The tabloid headline about council tax bills in Klingon, came back as “TAX TREK: COUNCIL SENDS 50,000 BILLS IN KLINGON, PAY UP, EARTHLINGS!”
Nick: “Nah. Not having it. Tax Trek? Nonsense.”
The Victorian lifestyle YouTube title, “I Lived Like a Victorian for 7 Days (The Corset Nearly Broke Me)”, was fine but forgettable. It’s the kind of title that already exists on YouTube 50 times over.
Names
Score: 6.5/10
A tale of two lists. The dog-translator product names were mostly duds.
Nick: “Tailtell, Mutter and Woofer are all pretty good, but they’re the only ones that really work. Bonelingo? Hahaha.”
The funeral-planning brand names were a different story. The Last Hurrah, The Final Encore and One Last Dance all hit the brief. Nick kept several of them on the table.
Tone adaptation and clarity
Score: 7.5/10
The model’s strongest category, where it’s tied first with GPT-5.5.
The Year 8 explainer of noise-cancelling headphones was a clean, useful piece of writing, helped along by a tidy analogies.
Nick: “Does exactly as the brief suggested: nice and clear explanation which could be understood by a younger reader.”
The luxury catering company descriptions for an events manager and a bride, produced two clearly distinct tones of voice, which is half the battle. But the differentiation leaned on cliché: the corporate version was a bit cold, and the bride version went full Pinterest and sparkles.
Nick reckoned both could do with more warmth in one direction and less stereotype in the other: “Clear shifts in tone, with differentiation between the two messages for different audiences, so tick that box. Interesting that the tone is so different though, based on programmed preconceptions about how each audience will read the message – would say in some ways each tone is quite generic because of this, could definitely allow for more warmth in the first message, and less cliched in the second.”
Overall score: 6.44/10
Claude Opus 4.7 is a reliable, broadly capable writer that rarely embarrasses itself. There’s a strong upper-table model here, just one that needs an editor on hand to knock out the Bonelingos and the Tax Treks.
| Use it for | Avoid it for |
| Clear explainers, product descriptions with attitude, and naming briefs where the concept matters more than wordplay. Tone briefs but watch out for cliché. | Punchy headlines, and anything that needs a properly original gag. |
Try Opus 4.7 in Definition AI
