OpenAI Launches GPT-6.1 Terra At Half The API Cost Of Sol - Dataconomy
OpenAI has added GPT-6.1 Terra to its range of general-purpose AI models. The company said its standard API prices are 50% lower than those of GPT-5.6 Terra and GPT-6.1 Sol.
The mid-range model costs $1 for every million input tokens and $5 for every million output tokens. OpenAI said these prices make Terra useful for tasks that run at a large scale, such as reading incoming invoices and entering their details into accounting software.
Artificial Analysis, a group that tracks AI model performance and costs, said GPT-6.1 Terra makes AI more cost-efficient. It described Terra as the first model in its class since GPT-5.6 and said it was “more intelligent for less than a third of the cost.”
OpenAI's general-purpose models are divided into four levels, from lower-cost models to more powerful ones. Luna is at the low end, Terra is in the middle, and Sol and Astra are at the high end. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens. GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens. GPT-6 Astra costs $10 per million input tokens and $50 per million output tokens.
Ted Sanders, a technical staff member at OpenAI, said in a comment on Hacker News that Terra is probably better than Astra and Sol at explaining technical topics. He said Terra is about as capable as those models at understanding images and using computers, but works much faster. However, he said it is somewhat less capable at most other tasks.
Sanders described Terra as an API option for developers who want a model that costs less than Sol but is much smarter than Luna, especially for understanding images, using computers and communicating clearly.
“For subscribers, it’s not a bombshell release,” Sanders said. He explained that Terra might help subscribers get more use out of their usage limits. It also gives them another option if they want clearer communication or faster computer use.
Terra's knowledge cutoff is June 1, 2026, which means it has the most recent knowledge cutoff in its model family. Astra and Sol have training-data cutoffs of April 30, 2026.
In safety tests, Terra responds to harmful requests slightly less often than GPT-6.1 Sol. It also refuses requests that it could safely answer less often than Sol does.
Data scientist Max Woolf argued in a recent blog post that the main change is economic: AI model prices keep falling while model performance improves. He noted that many people expected OpenAI and Anthropic to raise the prices of running their models or offer less value through subscriptions. These changes were expected to help bring their income closer to what they spend on training models, running them and maintaining the infrastructure they need.
Woolf said there was no strong evidence from behavioral economics or recent price drops for better models to support the idea that token prices would suddenly rise. He also pointed to price cuts by AI providers other than OpenAI and Anthropic. He argued that companies whose cost per token stays below the amount they earn per token will keep lowering prices to compete. This would put pressure on OpenAI and Anthropic to do the same.
Investors may soon be able to place bets on whether these pricing trends can support a sustainable AI business when Anthropic lists its shares on the Nasdaq stock exchange.