· ai-devtools
OpenAI releases GPT-6.1 Terra, a budget mid-tier model, priced 50 percent lower than GPT-5.6 Terra and GPT-6.1 Sol
The new GPT-6.1 Terra model from OpenAI offers strong capabilities while maintaining a lower price point.
OpenAI's general-purpose AI models are structured across four distinct tiers. The high-end offerings are labeled "Sol" or "Astra," carrying corresponding premium price tags. Conversely, the "Luna" models represent the most budget-friendly options. This newly introduced model utilizes the "Terra" designation, which the company reserves for its mid-range offerings.
OpenAI stated, "Terra continues to advance the cost-intelligence frontier across our model families, providing a new balance of speed, capability, and cost." Its standard API pricing is 50 percent lower than that of GPT-5.6 Terra and GPT-6.1 Sol, making it viable for developers automating high-volume tasks, such as processing incoming invoices and inputting their data into accounting software.
Artificial Analysis, a site tracking model performance and cost, points out that GPT-6.1 Terra significantly advances cost efficiency. The site reports that "It is the first Terra-class model since GPT-5.6, and is more intelligent for less than a third of the cost."
The pricing structure places GPT-6 Luna at the lowest end, with costs of $0.10/M input tokens and $0.50/M output. This is followed by GPT-6.1 Terra ($1/M, $5/M), GPT-6.1 Sol ($2/M, $10/M), and the top-tier GPT-6 Astra ($10/M, $50/M).
Ted Sanders, a technical staff member at OpenAI, detailed the model differences in a Hacker News comment. He noted that "In the case of GPT-6.1 Terra, I'd say Terra is probably better than Astra and Sol at technical communication, similar on vision and computer use (but much faster!), and a bit worse on most other things." He added that for API developers, "it's a nice midrange option if you want something cheaper than Sol but a lot smarter than Luna (especially for vision/computer use/communication)."
Terra's knowledge cutoff date is set to June 1, 2026, making it the most current in its family. This contrasts with Astra and Sol, whose training data cutoff was April 30, 2026. Furthermore, while Terra handles harmful requests less frequently than GPT-6.1 Sol, it also exhibits fewer unnecessary refusals than its more capable counterpart.
Max Woolf, a data scientist, highlighted the economic implications of these model trends in a recent blog post. He observed that the notion of sudden token price increases is "more a fantasy that is not supported by both behavioral economics and the actual trends of LLM prices going down significantly over the last few months for better models, and not just from OpenAI and Anthropic." He concluded that as long as providers can price below revenue per token, inference providers will drive down costs to remain competitive.
