A cheaper and more powerful ChatGPT: OpenAI’s new GPT-6.1 Terra model has found its recipe

Until then, OpenAI’s Terra model had remained on the GPT-5.6 generation (released last July). It has now moved to GPT-6.1 and become more powerful, but above all half the price.OpenAI considers this model capable of handling many different tasks, and it is already available.

GPT-6.1 Terra is here: what really changes?

On the page dedicated to this model, OpenAI explains that GPT-6.1 Terra is a so-called intermediate model (neither the cheapest nor the most powerful) designed to analyze what is displayed on your screen and then take action. Clicking, typing text, filling out a form, switching to another application… it can do almost anything. It therefore succeeds GPT-5.6 Terra, whose release was suspended by the Trump administration on grounds of national security. Only a select few partners had been able to test it. A necessary system, given that just a few days ago, two OpenAI AI agents carried out surprising actions. A few days later, the model was finally released.

GPT-6.1 Terra is designed for professional work, with a wide range of uses: writing an email, interpreting a technical diagram, extracting data from an invoice and entering it into accounting software, assisting with video editing, etc. OpenAI therefore presents it as a model combining detailed visual reasoning with precise computer use.

On the benchmark Agent’s Last Exam, this version performs 15% better than the previous version (on an exercise involving converting architectural plans into 3D models and analyzing an animated film). GPT-6.1 Terra is already available in ChatGPT Work and Codex. But for the time being, it is reserved for ChatGPT Plus, Pro, Enterprise and Edu subscribers.

An AI model twice as cheap as its previous version

Today, OpenAI’s model lineup has four tiers, from the least expensive to the most powerful: Luna, Terra, Sol and Astra. Terra is therefore in an intermediate position, between the most affordable model, Luna, and more expensive models such as Sol and Astra. GPT-6.1 Terra is announced at $1 per million input tokens and $5 per million output tokens. For reference, GPT-5.6 Terra was twice as expensive.

According to OpenAI’s pricing structure, this model is ten times more expensive than the least expensive model, namely GPT-6 Luna. However, it is ten times less expensive than the most powerful model, GPT-6 Astra, while offering comparable performance. OpenAI therefore appears to be positioning it as a model with the best possible performance-to-price ratio.

👉🏻 Follow tech news in real time: add 01net to your sources on Google and subscribe to our WhatsApp channel.