OpenAI has released GPT-6.1 Terra, the fast and cheaper tier of its current line-up, and the change most developers will notice is the bill. Standard API prices are $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens, which OpenAI says is half what GPT-5.6 Terra and GPT-6.1 Sol cost. For teams running agents that read invoices, fill in forms or click through desktop software all day, that halves the most predictable line in the budget.
OpenAI pitches Terra at office work: visual reasoning over drawings and screenshots, computer use and plainer writing. The benchmarks are its own selection. On Agents' Last Exam, a set of professional computer tasks, Terra at medium effort scores 15 percentage points above GPT-5.6 Terra at about a third of the cost, and OpenAI says it matches GPT-6 Astra at a tenth of the price per token. On Terminal-Bench 4.0 at maximum effort it solves 50.3% of tasks, against 21.5% for its predecessor. On an internal set of prompts designed to provoke factual mistakes, the error rate at low effort fell from 53.6% to 16.7%, and to 4.3% at maximum effort.
The comparison with Anthropic is where the claims get pointed. OpenAI says that at maximum effort Terra uses tokens more than three times as efficiently as Claude Opus 5.5 and costs a quarter as much per token, for roughly 9x lower cost per task. Our Opus 5.5 review lists $4 per million input tokens and $20 per million output, so the per-token gap is real; the efficiency half of the claim is OpenAI's measurement alone. OpenAI itself says GPT-6.1 Sol and GPT-6 Astra remain better for the hardest coding work, and on Terminal-Bench-Science it puts Terra within five points of Opus 5.5 rather than ahead.
Terra is available in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.1-terra. OpenAI rates it at the Critical threshold for cybersecurity under its Preparedness Framework, with the safeguards built for GPT-6 Astra, which is worth knowing for anyone deploying it with broad tool access.
The open question is how much of the gain survives outside OpenAI's own tests. Several benchmarks are internal, effort settings change the results sharply, and cost per task depends on how many tokens a model spends, which varies by workload. The price cut is certain. Whether Terra is the cheaper choice for a given job is something each team will need to measure on its own tasks.