Overview
- OpenAI announced GPT-6.1 Terra and made it available to ChatGPT Plus, Pro, Enterprise, Edu users and to developers through the OpenAI API under the model id gpt-6.1-terra.
- OpenAI’s internal benchmark results show Terra reduces factual-error rates versus the prior Terra generation and approaches the accuracy of higher-cost GPT-6 family models on visual-reasoning and professional tasks.
- Terra posts large gains on programming and automation tests, with Terminal-Bench 4.0 success rates and multi-step desktop-application performance both rising substantially compared with GPT-5.6 Terra.
- OpenAI published pricing for API use that positions Terra as a low-cost option (examples cited by the company include $1 per 1M input tokens and $5 per 1M output tokens) and says the model runs tasks faster and at far lower per-task cost than premium models and some competitors.
- OpenAI says Terra inherits safety measures from its premium models and improves prompt-injection resistance, though the company’s claims are based on its own internal tests and enterprises should validate performance, security, and limits before wide deployment.