I Tested Sonnet 5.5 vs Opus 5.5 vs GPT-6.1 Sol Inside a Recruiting Agency
Arbaz Surti | AI Automation
13 views
Claude Sonnet 5.5, Opus 5.5 and GPT-6.1 Sol | 200 resumes, one recruiting agency — I hid four findings in the pile, told none of the models they existed, and watched which ones found them.
A client wants a Senior Recruiting Coordinator. 200 people applied. Before anything ran, I wrote a sealed answer key with four things buried in that pool — none of them visible in any single resume. Then I gave all three models the same file, the same words, and the same medium setting, and asked each for two things: a dashboard I could run the requisition from, and a one-page brief I could send the client without editing it.
In this video you'll watch all six artifacts side by side and see the thing I didn't expect: the model that found the fewest findings still built the best tool, and the model that found all four built the dashboard I'd use least.
WHAT YOU'LL SEE:
✅ The sealed answer key, written and timestamped before a single model ran
✅ The exact prompt — word for word, identical for all three, naming no feature and no finding
✅ Sonnet 5.5's dashboard: the Meridian cluster, the pay-gap chart, light/dark mode
✅ GPT-6.1 Sol's dashboard: pipeline stages, drag-to-shortlist, a saveable working copy
✅ Opus 5.5's dashboard: how 200 became 1, and what to check on the call
✅ The ATS trap — the client's own two requirements quietly delete almost everyone good
✅ All three client briefs compared line by line, and which one I'd actually send
WHO THIS IS FOR:
→ Recruiting agency owners drowning in applications for one posted role
→ Small business owners trying to work out which AI model is worth paying for
→ Anyone who's been told the most expensive model is always the right one
→ Anyone still reading 200 applications by hand
You don't need a harness or a script. One prompt, pasted by hand into the tool you already have open.
💬 What's the biggest pile of documents sitting in your business right now? Drop it below.
⏱️ TIMESTAMPS:
00:00 - 00:48 - Three models, 200 resumes, four hidden findings
00:48 - 01:33 - The setup: why 200, and the two deliverables
01:33 - 02:18 - The sealed answer key
02:18 - 02:41 - The prompt, word for word
02:41 - 03:02 - Firing all three
03:02 - 07:08 - Sonnet 5.5's dashboard
07:08 - 11:08 - GPT-6.1 Sol's dashboard
11:08 - 13:29 - Opus 5.5's dashboard
13:29 - 14:12 - The ATS trap, and the filter that proves it
14:12 - 14:46 - Round 1 goes to GPT-6.1 Sol
14:46 - 15:06 - Round 2: the client brief
15:06 - 16:54 - Sonnet 5.5's brief
16:54 - 18:52 - GPT-6.1 Sol's brief
18:52 - 21:09 - Opus 5.5's brief, and the verdict
21:09 - 21:28 - Outro
🔗 RESOURCES & LINKS:
Claude AI: https://claude.ai
ChatGPT: https://chatgpt.com
Visual Studio Code: https://code.visualstudio.com
n8n (workflow automation): https://n8n.io
💬 YOUR TURN:
Three models, one prompt, three completely different tools. Which one would you have picked — the one that found everything, or the one that built the thing you'd actually open on Monday? Tell me below.
📬 STAY CONNECTED:
→ Subscribe for more AI automation tutorials
→ Hit the bell so you don't miss the next one
→ Share this with any recruiter still reading applications one at a time
#ClaudeAI #ClaudeCode #Sonnet55 #Opus55 #GPT6 #ChatGPT #AIAutomation #Recruiting #RecruitingAgency #HiringAutomation #SmallBusiness #BusinessOwner #AITools #Automation
1 Comments
I Tested Sonnet 5.5 vs Opus 5.5 vs GPT-6.1 Sol Inside a Recruiting Agency
Arbaz Surti | AI Automation
13 views
Claude Sonnet 5.5, Opus 5.5 and GPT-6.1 Sol | 200 resumes, one recruiting agency — I hid four findings in the pile, told none of the models they existed, and watched which ones found them.
A client wants a Senior Recruiting Coordinator. 200 people applied. Before anything ran, I wrote a sealed answer key with four things buried in that pool — none of them visible in any single resume. Then I gave all three models the same file, the same words, and the same medium setting, and asked each for two things: a dashboard I could run the requisition from, and a one-page brief I could send the client without editing it.
In this video you'll watch all six artifacts side by side and see the thing I didn't expect: the model that found the fewest findings still built the best tool, and the model that found all four built the dashboard I'd use least.
WHAT YOU'LL SEE:
✅ The sealed answer key, written and timestamped before a single model ran
✅ The exact prompt — word for word, identical for all three, naming no feature and no finding
✅ Sonnet 5.5's dashboard: the Meridian cluster, the pay-gap chart, light/dark mode
✅ GPT-6.1 Sol's dashboard: pipeline stages, drag-to-shortlist, a saveable working copy
✅ Opus 5.5's dashboard: how 200 became 1, and what to check on the call
✅ The ATS trap — the client's own two requirements quietly delete almost everyone good
✅ All three client briefs compared line by line, and which one I'd actually send
WHO THIS IS FOR:
→ Recruiting agency owners drowning in applications for one posted role
→ Small business owners trying to work out which AI model is worth paying for
→ Anyone who's been told the most expensive model is always the right one
→ Anyone still reading 200 applications by hand
You don't need a harness or a script. One prompt, pasted by hand into the tool you already have open.
💬 What's the biggest pile of documents sitting in your business right now? Drop it below.
⏱️ TIMESTAMPS:
00:00 - 00:48 - Three models, 200 resumes, four hidden findings
00:48 - 01:33 - The setup: why 200, and the two deliverables
01:33 - 02:18 - The sealed answer key
02:18 - 02:41 - The prompt, word for word
02:41 - 03:02 - Firing all three
03:02 - 07:08 - Sonnet 5.5's dashboard
07:08 - 11:08 - GPT-6.1 Sol's dashboard
11:08 - 13:29 - Opus 5.5's dashboard
13:29 - 14:12 - The ATS trap, and the filter that proves it
14:12 - 14:46 - Round 1 goes to GPT-6.1 Sol
14:46 - 15:06 - Round 2: the client brief
15:06 - 16:54 - Sonnet 5.5's brief
16:54 - 18:52 - GPT-6.1 Sol's brief
18:52 - 21:09 - Opus 5.5's brief, and the verdict
21:09 - 21:28 - Outro
🔗 RESOURCES & LINKS:
Claude AI: https://claude.ai
ChatGPT: https://chatgpt.com
Visual Studio Code: https://code.visualstudio.com
n8n (workflow automation): https://n8n.io
💬 YOUR TURN:
Three models, one prompt, three completely different tools. Which one would you have picked — the one that found everything, or the one that built the thing you'd actually open on Monday? Tell me below.
📬 STAY CONNECTED:
→ Subscribe for more AI automation tutorials
→ Hit the bell so you don't miss the next one
→ Share this with any recruiter still reading applications one at a time
#ClaudeAI #ClaudeCode #Sonnet55 #Opus55 #GPT6 #ChatGPT #AIAutomation #Recruiting #RecruitingAgency #HiringAutomation #SmallBusiness #BusinessOwner #AITools #Automation
What's the biggest stack of documents sitting in your business right now — resumes, invoices, contracts, intake forms? Tell me which one and I'll run it through all three next.