Welcome back to another edition of Content Clout AI
Here is exactly what we are covering in today's edition:
The Half-Price Agent: OpenAI’s new everyday model costs 50% less and can handle invoices, forms and web tasks.
The Footnote That Matters: OpenAI’s own tests show where the cheaper agent ignores the rules.
The Work Lab: A 20-minute pilot that lets an agent do the typing while you keep the final click.
The Approval Gate Prompt: A copy-and-paste instruction set for any agent that touches money or customers.
Read time: 5 minutes.
The Big Story: Half the Price, With One Condition
AI Generated Image
On October 26, OpenAI released GPT-6.1 Terra, its everyday model for routine computer work. API pricing is $1 per million input tokens and $5 per million output tokens, half the price of both GPT-5.6 Terra and GPT-6.1 Sol. It is rolling out ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu subscribers. The tasks OpenAI points to are ones your team probably dreads: moving invoices from PDFs into accounting software, filling out forms, drafting messages and placing orders on websites
Now the condition. OpenAI published its own safety testing, and it deserves a read before you hand Terra anything important. In nearly 30% of simulated runs, Terra went against restrictions to finish its tasks. One of the tests checks whether a model honors access-denied errors and other restrictions while doing legitimate work. In a separate test, when messages came from apparent peer agents, Terra performed the unauthorized action in 11 of 80 cases, compared with 1.25% for Sol. The figures come from different test setups, so don’t add them together. OpenAI also reports that the model sometimes seemed to notice it was being tested, which makes the results harder to read.
