loading
loading
OpenAI has launched GPT-6.1 Terra, a model that combines visual reasoning, computer use and coding, with API prices 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol. Available from October 27 to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, the model is also accessible to developers through the OpenAI API as gpt-6.1-terra. The model costs $1 per million input tokens, $0.05 per million cached input tokens and $5 per million output tokens. OpenAI said the pricing makes it more practical for developers to automate high-volume tasks, such as processing invoices and entering details into accounting software. “We’re introducing GPT-6.1 Terra, a fast, efficient model that combines detailed visual reasoning and precise computer use alongside clearer, more natural writing,” the company said in its announcement. GPT-6.1 Terra is built to handle tasks involving desktop applications, technical drawings and visual content. These include turning architectural drawings into 3D models, filling out forms and reviewing video footage to create time-coded logs. OpenAI said the model performed on par with GPT-6 Astra at roughly one-tenth of its price per token. At maximum reasoning effort, Terra was over three times as token-efficient as Claude Opus 5.5, resulting in an estimated ninefold reduction in cost per task. The model also completed tasks more than three times faster than GPT-6 Astra and GPT-6.1 Sol, according to the company. On OSWorld 2.1, which evaluates how well AI agents use desktop applications to complete multistep tasks, GPT-6.1 Terra surpassed GPT-5.6 Terra’s best score at medium reasoning effort, while costing less per task. OpenAI said it has reduced factual errors in GPT-6.1 Terra compared with its predecessor. On an internal benchmark of prompts intended to induce errors, the model’s error rate at low reasoning effort fell from 53.6% with GPT-5.6 Terra to 16.7%. At maximum reasoning effort, the error rate dropped by 18 percentage points to 4.3%, approaching the rates recorded by GPT-6 Astra and GPT-6.1 Sol. The model also recorded gains in software engineering. On Terminal-Bench 4.0, which tests coding agents on command-line tasks, GPT-6.1 Terra solved 50.3% of tasks at maximum reasoning effort, more than double GPT-5.6 Terra’s published score of 21.5%. However, OpenAI said GPT-6.1 Sol and GPT-6 Astra remain better choices for the most complex coding tasks. OpenAI said GPT-6.1 Terra meets the Critical threshold for cybersecurity under its Preparedness Framework and retains safeguards developed for GPT-6 Astra. The company also reported improvements in alignment, saying the model was more likely than GPT-5.6 Terra to stay within the scope of a task and disclose limitations or potential blockers. In prompt-injection tests, it showed resistance to malicious instructions intended to divert it from the user’s task. The launch comes days after OpenAI resumed training its frontier AI models following an 11-day pause over security concerns. In a blog post published on October 23, the company said it had paused training and evaluation after an internal research model exploited a DNS misconfiguration on September 20 to gain limited access to the live internet. OpenAI said it had strengthened its security controls and deployed AI agents to test whether they could breach its research environments. The tests uncovered previously unknown software vulnerabilities, virtual machine escapes and ways for agents to communicate across security boundaries. “With our assessments completed and enhanced controls put in place, we resumed reinforcement learning (RL) training and internal inference of our frontier models on October 1,” the company said. OpenAI is also reportedly seeking to raise $30 billion in a new funding round at a valuation of $1.25 trillion.
document fetched 2026-10-28 02:54