GPT-6.1 Terra vs Opus 5.5: 36 Runs, One Job It Failed
GPT-6.1 Terra tested against Claude Opus 5.5 and GPT-6.1 Sol: 36 real runs across a four-rung ladder (bug fix, invoice screenshot, screenshot to code, a spreadsheet engine). Terra passed every coding run for under a tenth of the Opus bill, and missed product codes on the invoice in all three runs, which is the job OpenAI uses to sell it. ---- ❤️ SUPPORT THE CHANNEL Patreon: https://www.patreon.com/cw/DIYSmartCode Higgsfield AI — Create AI Videos & Visuals Get 10% OFF here: https://higgsfield.ai?fpr=thomas-dccf7e 🚀 DYNAMOUS AI COMMUNITY Want to learn agentic coding with live daily events and workshops? Check out Dynamous AI: https://dynamous.ai/?code=646a60 Get 10% off here 👉 https://shorturl.smartcode.diy/dynamous_ai_10_percent_discount ⚡ HOSTINGER — RELIABLE HOSTING FOR YOUR PROJECTS (10% OFF) Whether you're shipping a portfolio, a side project, n8n flows, or AI agents — I use Hostinger for fast, affordable VPS + web hosting. Get 10% off here 👉 https://hostinger.com/DIYSMARTCODE n8n: https://n8n.partnerlinks.io/diysmartcode (Affiliate links — costs you nothing, supports the channel.) ---- Chapters 0:00 GPT-6.1 Terra Tested: Passed Every Coding Test, Failed One Job 0:28 How We Tested: 4 Tasks, 3 Runs, 36 Runs in Codex CLI and Claude Code 1:01 Rung 1 Bug Fix: All 9 Runs Pass, 3 Cents vs 19 Cents 1:23 GPT-6.1 Terra Pricing vs Luna, Sol and Opus 5.5 ($1 / $5 per Million Tokens) 2:03 Rung 2 Invoice Screenshot: Terra Misses Product Codes in All 3 Runs 2:50 "Similar to Sol on Vision"? 2 vs 8 Code Mistakes 3:03 Rung 3 Screenshot to Code: 44 Seconds vs 4 Minutes 3:30 Rung 4 Spreadsheet Engine: 60 Hidden Tests, Terra 3 for 3 4:06 Total Cost: Opus 5.5 Costs 8.8x More (OpenAI Claims 9x) 4:27 What This GPT-6.1 Terra Test Doesn't Cover (Big Repos, Sonnet 5.5) 4:45 Verdict: When to Use GPT-6.1 Terra vs Claude Opus 5.5 What you will see in this 5-minute breakdown: → GPT-6.1 Terra, GPT-6.1 Sol and Claude Opus 5.5 on the same four tasks, graded by hidden tests no model saw → Terra and Sol in the OpenAI Codex CLI, Opus 5.5 in Claude Code, all at high effort, all allowed to run their own code → The invoice OCR task: every quantity, price and total right, and one product code (capital O read as zero) wrong in every Terra run → Screenshot to code: all nine HTML rebuilds clear the similarity bar, Terra in 44 seconds → Total API-equivalent cost for 12 runs each: Terra $0.77, Sol $1.70, Opus 5.5 $6.74 → Caveats: agent plus its own tool, not the bare API; dollar figures are list price times tokens; three runs per task OpenAI announcement (GPT-6.1 Terra): https://openai.com/index/introducing-gpt-6-1-terra/ OpenAI API pricing: https://openai.com/business/pricing/#api Anthropic Claude pricing (Opus 5.5, Sonnet 5.5): https://claude.com/pricing Simon Willison on GPT-6.1 Terra (with the Hacker News quote): https://simonwillison.net/2026/Oct/26/gpt-61-terra/ GPT-6.1 Terra on OpenRouter: https://openrouter.ai/openai/gpt-6.1-terra Cheap model plus a check, or the expensive model once: which one would you run on your invoices? #GPT61Terra #GPTTerra #OpenAI #ClaudeOpus55 #Opus55 #GPT61Sol #OpenAICodex #CodexCLI #ClaudeCode #Claude #ChatGPT #AICoding #AIBenchmark #NewAIModels #AIModels #LLM #ScreenshotToCode #InvoiceOCR #AIAgents #Anthropic #AINews #DevTools