OpenAI filled the gap in its GPT-6 line on Monday with GPT-6.1 Terra, a mid tier model built for computer use, visual reasoning, coding and everyday writing. It is available through the API as gpt-6.1-terra and in ChatGPT Work and Codex for paid plans, and Amazon Web Services made it available on Bedrock the same day. The headline is price: $1 per million input tokens and $5 per million output tokens, about half the price of GPT-5.6 Terra and of GPT-6.1 Sol, and one tenth of the flagship Astra’s $10 and $50 rates.
The performance claims are OpenAI’s and are specific. On Agents’ Last Exam, its benchmark of professional computer tasks, the company says Terra scores 15 percentage points above GPT-5.6 Terra at about a third of the cost per task, performs on par with Astra on that test, and finishes tasks more than three times faster than Astra or Sol. Its error rate at low reasoning effort fell from 53.6 percent to 16.7 percent on an internal error inducing prompt set. Independent scoring by Artificial Analysis puts Terra at 48 on its Intelligence Index, six points above its predecessor, at under a third of the cost per task, while noting the strongest claims rest on the single Agents’ Last Exam benchmark.
The positioning matters more than the benchmark. Terra is aimed at the unglamorous middle of office work, reading screens, filling forms, entering invoices, drafting documents, which is where enterprise token volume actually sits. Yesterday’s edition carried Aiven cutting its internal token bill by about 80 percent by routing work away from frontier models, and Apollo’s argument that mainstream firms treat token spending as a cost centre. A flagship class computer use model at mid tier prices is OpenAI’s answer to both: it lowers the cost per task before a gateway or a rival does it for the customer.
OpenAI still points the hardest coding and research work to Sol and Astra. What to watch is whether mid tier pricing holds across the industry: Anthropic, whose S-1 leads today’s edition, and Google both sell into the same agent workloads, and a halving at this tier resets the comparison for every enterprise renewal this quarter.