GPT-6.1 Terra cuts API costs, widening the budget for revisions

OpenAI has released GPT-6.1 Terra, a mid-range model aimed at everyday office work and automated software control, according to the-decoder.com. Its API rates are $1 per million input tokens and $5 per million output tokens, compared with $2 and $12 for GPT-5.6 Terra. OpenAI says its intended tasks include drafting proposals, interpreting technical drawings and filling out forms.

For teams budgeting a drafting workflow, the useful comparison is how much processing those rates buy. Hypothetical calculation: a batch consuming 10 million uncached input tokens and 2 million output tokens would cost (10 × $1) + (2 × $5) = $20 on GPT-6.1 Terra. The same token volumes on GPT-5.6 Terra would cost (10 × $2) + (2 × $12) = $44. These are token-charge calculations, not measured task costs.

That creates a concrete budget choice: two Terra passes at those hypothetical volumes would cost $40, still below one predecessor pass at $44. A team could therefore consider a drafting-and-revision workflow within its previous token budget. The condition matters: this assumes each pass consumes the stated volumes. Longer revision prompts, additional output or further retries change the total, and a second pass does not establish that the result is correct. The decision is whether the revised output meets the workflow’s acceptance criteria within that $44 ceiling.

The reported benchmarks support considering Terra for more work, but do not settle which tasks to move. Artificial Analysis reports a Coding Agent Index v2.0 score of 48 for Terra, below GPT-6.1 Sol’s 54. OpenAI still recommends Sol and GPT-6 Astra for the most demanding coding tasks. Taken together, those facts suggest a selective routing decision: consider Terra where the lower processing budget is useful and acceptable output can be established, while retaining the larger models for demanding coding work. The calculation establishes budget headroom; it does not establish equivalent quality.