A free AI writing model test that hides the names of 32 AI models while you judge their drafts on 18 creator writing jobs, then shows who wrote your favorite and what it cost.
Judge drafts from 32 AI models with their names and prices hidden, and find the one that writes best for you.
This free AI writing model test shows you which AI writes best for you, without the brand name getting a vote.
You judge two drafts at a time, written by 32 AI models from the same brief, and pick the better one.
Here's how the test works, the jobs and their briefs, and what to do with your result.
TL;DRKey takeaways
What is an AI writing model test?
An AI writing model test is a blind comparison: several AI models write the same piece, and you judge the drafts without knowing which model wrote which.
It shows your own taste on your own kind of writing, which a benchmark score can't.
How the AI writing model test works
Each job puts 10 of its drafts, picked at random, through 9 duels:
- Two drafts appear side by side, with no names and no prices.
- You pick the better one, and it stays in to meet the next draft.
- The draft left standing after every duel is your favorite for that job.
The two drafts swap sides at random every duel, so the side a draft sits on tells you nothing. Across everyone's runs, every model gets judged on every job.
When you finish, the names come back. You see what each draft cost per 1,000 words, and how many seconds it took to write for the models I ran through their API.
Pick more than one job, and your ranking adds up your choices across all of them.
The 18 writing jobs
Every model got the same 18 jobs a creator does every week:
| Job | What each model wrote |
|---|---|
| YouTube video intro | The first 60 seconds of a YouTube script, spoken to camera |
| YouTube title and description | Three title options and the video description |
| Newsletter issue opening | A subject line, preview text and the issue's opening |
| Newsletter issue, full | A full newsletter issue built on one idea |
| X thread | An X thread of 8 to 10 posts |
| LinkedIn post | A first-person LinkedIn post written from notes |
| Instagram carousel | An Instagram carousel, slide by slide, and its caption |
| Instagram Reel script | A 30 to 40 second Reel script with shots and on-screen text |
| Blog post intro and outline | SEO title, meta description, intro, key takeaways and outline |
| Blog post, full | A full how-to blog post, ready to publish |
| Course sales page section | A course sales page section: who it's for, what's inside, guarantee, button |
| Product launch email | A launch email: subject, preview and body |
| Webinar invite email | An invite email for a free live workshop |
| Sponsor pitch email | A cold email pitching a brand on a sponsored video |
| Podcast guest pitch email | A cold email pitching yourself as a podcast guest |
| Podcast show notes | Show notes: title, summary, timestamps, quote and links |
| Customer support reply | A reply to an angry refund request outside the refund window |
| Creator bio | A website About section, a social bio and a speaker bio |
Each brief gives every fact a draft may use. So when a model makes up a number or a result, you can spot it.
Open "Read the brief" during the test to see the exact words every model got.
The models in the test
The test has the newest model you can use from each major AI company as of October 1, 2026, plus a cheaper sibling where there is one.
They come from OpenAI, Anthropic, Google, xAI, Meta, DeepSeek, Qwen, Moonshot AI, Z.ai, Xiaomi, Mistral, MiniMax, Tencent, NVIDIA, Cohere, ByteDance Seed and Thinking Machines, plus Jev Router from TypeSafe, which picks a model for each job itself.
Every model got the same instructions at low reasoning effort, with two exceptions. Command A+ ran with reasoning off, because at low effort it never finished a draft, and Jev Router picks its own model and effort.
The drafts were written once, so every visitor judges the same ones.
Everyone's votes build a public ranking
Every choice you make is saved as an anonymous vote, with no name or email attached.
The votes rank the models on the best AI model for writing board, the way blind voting ranks image and video models.
Your own result shows your taste. The board shows what most people prefer.
How to use your result
Once you know which AI writes best for you:
- Use it for your first drafts, and keep your edits for the voice only you have.
- Check the price, since a cheaper model that came close saves money on every draft.
- Run your own brief through your top two before you pay for a plan.
Then run your drafts through the AI slop checker to find anything that still reads as AI.
Who is the AI writing model test for?
It's for anyone who writes with AI and wants to pick a model on quality, not brand:
- AI prompt generator – creators who want sharper briefs for whichever model wins
- Headline analyzer – bloggers checking the titles their AI writes
- Email subject line tester – newsletter writers testing subject lines before a launch
- AI humanizer – anyone who wants a draft to sound more like them
Start with the test, then let these tools polish what your favorite model writes.
Try your winner in your own AI
Open the model that won your test in its own app, and give it your next brief. Some links are affiliate links.
Chat apps
FAQs about the AI writing model test
Questions about the AI writing model test? Here's what to know.
It's a blind test where AI models write the same piece, and you pick the better draft without knowing which model wrote it.
The one you prefer on your own kind of writing, which is what this test finds.
The public board shows which model most people prefer.
Each job puts 10 of its drafts, picked at random, through 9 duels.
You judge two drafts at a time, and the one you pick meets the next draft.
After the last duel, the draft left standing is your favorite.
Yes: every model wrote its drafts from the same brief, through its maker's own app or OpenRouter, and nothing was edited.
A reply that wasn't a draft at all, like a model's own planning notes, is left out of the duels.
A name or a price changes what people pick, so both stay hidden until you finish.
The newest models from each major AI company as of October 1, 2026, 32 in all, from OpenAI, Anthropic, Google, xAI, Meta, DeepSeek, Qwen, Moonshot AI, Z.ai, Xiaomi, Mistral, MiniMax, Tencent, NVIDIA, Cohere, ByteDance Seed and Thinking Machines, plus Jev Router from TypeSafe.
It's what the draft cost at the model's API price, reasoning included, scaled to 1,000 words.
For the models I ran on my chat plans, it's the price of the same tokens through the API.
Chat apps charge a monthly plan instead, so use it to compare models, not to budget a subscription.
Your result is your taste on the jobs you picked.
The board adds up everyone's votes on every job.
Yes: a vote stores which draft won and which lost, with your browser's random id and your IP address hashed, never your name or email.
Yes: it's free with no sign-up.
It never calls an AI while you use it, since the drafts were written ahead of time.
I did. I'm Navid Moazzez, and I made the AI writing model test as one of my free tools on navid.me.
More about me.
Navid Moazzez
AI business strategist & AI OS builder
Navid Moazzez helps creators and founders master AI and build their own AI Operating System (AI OS) to automate their business and life.
Navid.me is reader-supported. When you buy through links on this site, I may earn an affiliate commission. Learn more.
More free tools
X
Bluesky
LinkedIn
Threads
Mastodon
InstagramRelated MCP servers & CLIs
Substack
TikTok
Apple Podcasts
Free AI newsletterThe most actionable AI newsletter for founders
Every week, get proven AI strategies, curated tools, and step-by-step systems to grow your audience, create better content, and build a profitable creator business.
No fluff, no filler, no BS. Just five minutes each week that might level up your online business and life.
P.S. Sign up now to get free access to my ultimate AI tools guide for creators.
