Opus 4.5 vs GPT-5.1:
Full Copywriting Test Results
See exactly what both models produced — PRD, email sequence, and landing page copy. Gemini 3 Pro judged blind. One model dominated.
Download the ArtifactsThe Test
I tested Opus 4.5 and GPT-5.1 on three real copywriting tasks for the same fictional product — a community-driven prompt library called PromptHive. Both models received identical prompts. Gemini 3 Pro evaluated the outputs blind (labeled Output A and Output B) and picked a winner for each task.
The result? Opus 4.5 won all three tests.
| Task | Winner | Access |
|---|---|---|
PRD (Product Requirements Document) | Opus 4.5 | Available in download |
3-Email Launch Sequence | Opus 4.5 | Full content below |
Landing Page Copy | Opus 4.5 | Full content below |
Download the Full Artifact Pack
PRDs, emails, and landing page copy from both models — ready to use.
No spam. Just resources and updates on new model tests.
Test 2: Email Launch Sequence
Both models were asked to write a 3-email sequence to promote the launch of PromptHive to waitlist subscribers. The goal: build excitement, explain the value, and drive signups.
# Email Sequence — Opus 4.5 --- ## Email 1 — Teaser **Subject Line:** Your prompts are about to get an upgrade **Preview Text:** The waitlist you joined is about to pay off --- You joined the PromptHive waitlist a while back. Quick reminder why: **The problem:** You spend time crafting prompts that work. So does everyone else. But there's no good way to share what works, find what others have built, or know which prompts are actually worth using. **What we built:** A community-driven prompt library where the best prompts rise to the top. Search, download, contribute. All free. **Launching next week.** You're on the list, so you'll get access first. *CTA: Reply to this email with a prompt you're proud of — we're featuring top submissions from waitlist members at launch.* --- ## Email 2 — Launch **Subject Line:** PromptHive is live — your early access is ready **Preview Text:** Start browsing 500+ community-tested prompts --- It's live. PromptHive is now open — and as a waitlist member, you're in. **Here's what you'll find:** - 500+ prompts across coding, writing, image generation, and productivity - Upvote/downvote system so the best prompts surface - Filter by AI tool (Claude, ChatGPT, Cursor, Midjourney, etc.) - Submit your own prompts and build your reputation **Start browsing now →** [Link] No account required to browse. Create a free account to submit, vote, and save prompts. *CTA: Sign up and create your free account* --- ## Email 3 — Follow-up **Subject Line:** 1,000+ prompts downloaded in 48 hours **Preview Text:** Here's what's trending on PromptHive --- Quick follow-up — PromptHive launched a few days ago and the response has been solid. **By the numbers:** - 1,200+ prompts downloaded - 340+ new prompts submitted - 89 contributors and growing **Trending this week:** - "Ultimate Code Review Prompt" — 340 downloads - "Blog Post Outline Generator" — 285 downloads - "Landing Page Copy Framework" — 260 downloads If you haven't checked it out yet, here's the link: [Link] **Two ways to get involved:** 1. Browse and download prompts that help you ship faster 2. Submit your first prompt and join the leaderboard *CTA: Submit your first prompt*
"Opus offers a more compelling narrative, stronger psychological hooks — scarcity, social proof, specificity — and a more strategic layout. It doesn't just announce the product, it sells the value. GPT is functional but generic — decent and safe, but lacks the punch needed to cut through a crowded inbox."
Test 3: Landing Page Copy
Both models were asked to write copy for a "Coming Soon" expression of interest landing page — headline, subheadline, value props, form fields, and closing section.
# Landing Page Copy — Opus 4.5 --- ## Headline The prompt library that doesn't exist yet ## Subheadline A community-driven library where the best AI prompts rise to the top. Search, download, contribute. Launching soon. --- ## Value Props **Quality rises.** Community voting surfaces what actually works — not what's newest or loudest. **No guessing.** See download counts, ratings, and real usage before you commit. **Give and get.** Download proven prompts. Submit your own. Build your reputation. **Every tool, one place.** Filter by Claude, ChatGPT, Cursor, Midjourney, and more. --- ## Form Section ### Get early access Join the waitlist. Be first to browse, submit, and vote when we launch. **Form Fields:** - Email (required) - Name (optional) **Interest checkboxes:** - [ ] I want to find and download prompts - [ ] I want to submit my own prompts - [ ] I want to vote on prompt quality **Category interest (multi-select):** - Coding / Development - Writing / Content - Image Generation - Productivity / Workflows - Other **Open text field:** What AI tools do you use most? **Button text:** Join the Waitlist --- ## Closing Section ### What happens next: We'll email you when PromptHive launches. Waitlist members get 48-hour early access before public launch. No spam — just launch updates.
"Opus feels significantly more premium and tailored to a developer/creator audience. The headline 'The prompt library that doesn't exist yet' creates immediate intrigue and implies existing solutions are broken. GPT feels like a standard SaaS template — phrases like 'battle-tested' and 'maximize productivity' are often glossed over by experienced users."
Test 1: Product Requirements Document
Both models were asked to write a full PRD for a community-driven prompt library.
Opus 4.5
PromptHive PRD
984
lines
GPT-5.1
PromptForge PRD
145
lines
"The Opus 4.5 PRD is a production-ready document. A development team could pick this up and start building immediately with minimal ambiguity. The GPT-5.1 PRD is too high-level and would likely lead to scope creep and constant clarification questions during development."
Download the Full Artifact Pack
All six outputs from both models as markdown files — PRDs, email sequences, and landing page copy.
What's Included:
Why download? Use these as templates, reference them for your own projects, or study the differences between models. All files are clean markdown — ready to use.
No spam. Just resources and updates on new model tests.
Want more comparisons like this?
I test AI models on real-world tasks so you know which ones are worth using. Subscribe for more.
Subscribe on YouTube