GPT-6.1 Terra: OpenAI's Practical Midrange Model

OpenAI unveiled GPT-6.1 Terra, positioning it as a midrange model equivalent to Anthropic’s Sonnet within its product lineup. The company emphasized the model’s visual reasoning and computer use capabilities alongside improved natural writing. Terra builds on GPT-5.6 Terra, handling everyday business workflows like proposal drafting, technical drawing interpretation, form completion, and video editing.

Terra is priced at $1 per million input tokens and $5 per million output tokens—substantially less expensive than both GPT-6.1 Sol and GPT-5.6 Terra but considerably more costly than GPT-6 Luna. This positioning makes it useful for developers seeking better performance than Luna without paying for Sol, especially for vision-related and communication tasks. Ted Sanders from OpenAI noted Terra excels in technical writing compared to Astra and Sol, matches their vision and automation capabilities while being faster, though falls short on other dimensions.

For API developers, Terra provides a practical middle ground. For subscription users, the release offers limited appeal unless they’re approaching usage limits or specifically need faster computer automation. The model’s lower cost particularly benefits use cases involving repeated automation tasks where less expensive processing can run continuously.