OpenAI's mid-tier Terra drops 50% in price with GPT-6.1 update

OpenAI has refreshed Terra, the model that anchors the middle of its lineup. Terra was left out last month when the company announced Sol and Luna as part of the GPT-6 family. The company has now closed that gap, moving Terra from GPT-5.6 to GPT-6.1.

The new model emphasizes speed and the ability to operate a computer, and it arrives at a lower price than its predecessor. GPT-6.1 Terra is available starting today through ChatGPT Work, Codex and the API.

Where Terra sits in the four-model lineup

OpenAI currently has four main model tiers. Ordered from lowest to highest in price and capability, they are Luna, Terra, Sol and Astra. Terra sits squarely in the middle of that structure.

According to the announcement, GPT-6.1 Terra delivers detailed visual reasoning and precise computer use, along with clearer and more natural writing. OpenAI describes the model as a notable step up from GPT-5.6 Terra and says it is strong at the communication and computer-based tasks that fill a working day. Examples given include drafting proposals, interpreting technical drawings, filling out forms and editing video.

Half the price, far fewer errors

One of the most striking details is the price. OpenAI says GPT-6.1 Terra costs 50 percent less than GPT-5.6 Terra, the model it replaces. That makes it an attractive option for anyone looking to automate computer-based work at lower cost.

Accuracy has improved as well. On OpenAI's internal benchmark of prompts prone to producing errors, the error rate at low reasoning effort fell from 53.6 percent for GPT-5.6 Terra to 16.7 percent for GPT-6.1 Terra, a drop of roughly 37 points. At the highest effort setting, the error rate fell by 18 points to 4.3 percent. According to the company, that figure is nearly on par with GPT-6 Astra and GPT-6.1 Sol.

Availability

GPT-6.1 Terra launches today in ChatGPT Work, Codex and the API, so the new mid-tier model is accessible on all three channels from day one.

By promising speed, computer use and lower cost together, the update offers an interesting option for those who want a model that can do more work for less money. The figures on price and error rates come from OpenAI's own statements; how the model performs in practice will become clearer as users test it in their own workflows.