OpenAI Launches GPT-6.1 Terra: Halves API Price, Matches Astra on Computer Use
OpenAI has released the mid-tier model GPT-6.1 Terra, focusing on 'Computer Use' capabilities. It scores approximately 60% on the Agents' Last Exam benchmark, matching the top-tier GPT-6 Astra. Available to Plus and higher subscribers at no extra cost, its API pricing is halved compared to the previous generation, offering efficient automation for daily tasks at a lower price point.
Full Report
OpenAI Releases GPT-6.1 Terra: Included in Plus and Higher Tiers at No Extra Cost, API Prices Halved, Computer Use Capabilities Match GPT-6 Astra
OpenAI has released GPT-6.1 Terra, with a primary focus on Computer Use: the AI observes screen content, clicks the mouse, types, and switches between applications autonomously, completing tasks like a human.
In OpenAI's model lineup, it sits in the middle tier—cheaper than GPT-6.1 Sol and the top-tier GPT-6 Astra, but more capable than the most affordable GPT-6 Luna.
GPT-6.1 Terra is currently rolling out to ChatGPT Plus, Pro, Enterprise, and Education users, and is selectable within ChatGPT Work and Codex. Free users are not included in this rollout, and the Business tier was not mentioned by official sources.
The most tangible benefit for subscribers is increased quota efficiency. According to official statements, using Terra in Work and Codex allows users to accomplish significantly more work per unit of quota compared to the previous GPT-5.6 Terra, and vastly more than when using Sol or Astra, though specific multipliers were not provided.
For developers calling gpt-6.1-terra via the API, pricing is set at $1 per million input tokens and $5 per million output tokens. OpenAI claims this is half the cost of GPT-5.6 Terra (third-party price trackers recorded the old rates at $2 and $12). For comparison, at the same unit scale, Sol costs 10, and Astra costs 50.
Advantages Over Previous Versions, Sol, and Astra
OpenAI benchmarked against Agents' Last Exam, a test suite evaluating an AI agent's ability to effectively utilize various software tools in professional workflows. According to charts published by OpenAI, Terra scores approximately 60% at its highest setting, while the previous GPT-5.6 Terra scores around 50%, and GPT-6 Astra scores about 59%, making them roughly equivalent. On a per-token basis, Terra's price is one-tenth that of Astra; officials also stated that Terra completes tasks on average more than 3x faster than both Astra and Sol.
On the same chart, the highest score belongs to Anthropic's Claude Opus 5.5 at approximately 63%, which also incurs the highest cost—roughly 9 times that of Terra's highest setting according to the graph.
OpenAI also acknowledged Terra's limitations: for deep research or complex coding tasks, Sol or Astra remain the recommended choices. For daily workflows involving clicking, filling forms, and writing across various applications, switching to Terra can save both quota and time.
Original Source: https://t.co/H4Uaw49Oay
↳ Quote @OpenAI (OpenAI): GPT-6.1 Terra: delivering fast and capable computer use at a lower price.
GPT-6.1 Terra understands what’s on screen, takes action across apps, writes naturally, and is a more capable model for your whole workday with the efficiency to take on more of it.
