OpenAI rolls out budget-priced mid-tier GPT-6.1 Terra
OpenAI has added GPT-6.1 Terra, a mid-tier model positioned between its cheaper Luna models and higher-priced Sol and Astra offerings. It matters to developers because it aims to deliver stronger capabilities than Luna at a lower cost than Sol, especially for high-volume tasks.
Terra costs $1 per million input tokens and $5 per million output tokens, half the listed prices of GPT-6.1 Sol and GPT-5.6 Terra, and a tenth of GPT-6 Astra’s price. OpenAI staff say it is particularly strong in technical communication and fast at vision and computer-use tasks; its knowledge cutoff is 1 June 2026, later than the 30 April cutoff for Astra and Sol. The article also points to falling model prices alongside rising performance, while noting that this trend leaves questions about the industry’s long-term economics.
In short
- GPT-6.1 Terra adds a lower-cost mid-tier option.
- It costs one tenth as much as GPT-6 Astra.
- Falling prices raise questions about AI business economics.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
A lower-cost model with stronger capabilities than the cheapest options could make advanced AI more accessible to developers and support high-volume uses that were previously too expensive. If performance keeps improving as prices fall, that may spur useful products and broader experimentation.
The case against
A cheaper model may still encourage greater dependence on a small number of AI providers, while putting pressure on the economics needed to sustain continued investment. Developers may also need to weigh the advertised price against whether Terra’s capabilities and reliability suit their particular tasks.