Franklin AI News Brief

GPT-6.1 Terra: API prices and vendor-reported task gains

Key Takeaways

  • OpenAI prices GPT-6.1 Terra at $1 input and $5 output per million tokens.
  • Vendor-reported task gains warrant tests of accepted results, retries and cost.
  • GPT-6.1 Terra targets visual reasoning, computer tasks and everyday writing at lower standard API prices.
  • OpenAI’s reported evaluation gains still need checks against a team’s own accepted-output costs.

OpenAI has introduced GPT-6.1 Terra for visual reasoning, computer use and everyday writing. Its announcement positions the model as a faster, less expensive option for work such as interpreting drawings, filling forms and processing invoices, alongside improvements over GPT-5.6 Terra.

Who can access GPT-6.1 Terra, and what are API prices?

The announcement says Terra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex. Developers can access it through the API as gpt-6.1-terra.

Standard API prices are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. OpenAI describes the standard prices as 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. A lower token price does not establish the total cost of a particular workflow, which also depends on how much work the model performs.

Which task gains does OpenAI report for Terra?

On Agents’ Last Exam, OpenAI reports a 15-percentage-point improvement over GPT-5.6 Terra at medium reasoning effort, at about one-third the cost. For Terminal-Bench 4.0 at maximum effort, it reports 50.3% of tasks solved, compared with the older model’s published 21.5% score.

These are vendor-reported benchmark results. OpenAI still identifies GPT-6.1 Sol and GPT-6 Astra as better choices for the most complex coding work. The appropriate comparison depends on the task and reasoning setting rather than the model name alone.

What should a Terra migration test measure?

OpenAI reports that errors on its internal error-inducing benchmark fell from 53.6% to 16.7% at low reasoning effort. That measures one evaluation setup and does not establish an error rate for arbitrary business documents.

Teams considering a switch can test representative tasks with known answers, inspect failure cases and measure completion cost. Computer actions also deserve checks of scope and intended changes, even when a model performs well on a benchmark.

Our read

Franklin AI Take

Terra’s standard API prices are $1 per million input tokens and $5 per million output tokens. That makes a migration test worth measuring, especially for repeated document or desktop tasks. Compare the cost of an accepted result, including retries and human corrections, rather than assuming the token-price reduction carries straight through to the final workflow.