OpenAI introduces GPT-6.1 Terra, a new model offering advanced capabilities at a lower cost, perfect for enterprises.
OpenAI has launched its latest model, GPT-6.1 Terra, designed to provide advanced reasoning skills at a significantly reduced cost. This innovative tool is tailored for organizations looking to enhance operational efficiency and accuracy in various professional tasks without breaking the bank.
A New Benchmark for Professional Work
GPT-6.1 Terra replaces the earlier GPT-5.6 Terra model and claims to deliver a 50% cost reduction while offering improved performance. It is specifically crafted to assist with routine tasks such as drafting documents, interpreting technical drawings, completing forms, and even video editing.
“With GPT-6.1 Terra, we’re providing a model that is not only fast and efficient but also excels in visual reasoning and user-friendly writing,” said OpenAI representatives.
Targeting Organizations for Better Automation
This new mid-tier model stands between OpenAI's lower-cost GPT-6 Luna and its high-end options like GPT-6.1 Sol and GPT-6 Astra. The company's focus on computer usage capabilities is supported by various benchmark scores that highlight its strengths in executing professional tasks.
According to Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting, “Enterprises require AI models that strike a balance between intelligence, speed, and affordability. The Terra model effectively fills this void, competing with alternatives like Anthropic’s Claude Sonnet and Google’s Gemini Flash.”
Promising Performance and Cost Savings
There's no doubt that GPT-6.1 Terra presents impressive performance metrics. In the Agents’ Last Exam benchmark, it scored 60.0%, with a cost of just $1.01 per task. This shows a considerable improvement compared to its predecessor, GPT-5.6 Terra, which managed 50.7% at a higher price of $3.58. For context, GPT-6.1 Sol scored 58.6% at $2.00 and GPT-6 Astra reached 59.3% at $7.24.
When it comes to visual reasoning on the MMMU-Pro benchmark, GPT-6.1 Terra achieved an 85.5% score at a mere $0.012 per task, outperforming GPT-5.6 Terra while closely rivaling the more expensive frontier models. This aspect emphasizes how enterprises can access high-quality performance without the exorbitant costs.
Reducing Errors and Improving Security
OpenAI has also prioritized minimizing hallucinations and factual inaccuracies within its outputs. The error rate from low reasoning efforts significantly dropped from 53.6% with GPT-5.6 Terra to 16.7% with the new model. Even at maximum effort, the rate decreased to 4.3%, making it comparable to the high-end options in the lineup.
In terms of cybersecurity, GPT-6.1 Terra meets essential safety standards and retains all the protective measures introduced in previous models, ensuring that companies can utilize the tool with confidence.
A Practical Choice for Enterprises
“GPT-6.1 Terra is particularly suited for high-volume tasks like invoice processing and document analysis,” Jain added. He further noted that while it excels in speed and accuracy, GPT-6.1 Sol remains the better choice for more complex tasks such as research and software development.
However, Jain advises caution. Relying solely on technical specification benchmarks is inadequate. Companies should assess how GPT-6.1 Terra performs under their specific workflows, paying attention to factors such as task completion rates, accuracy, duration, and compliance with security protocols.
Pricing Structure and Accessibility
GPT-6.1 Terra is readily available to Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex. Developers can also access it using the OpenAI API.
The pricing for GPT-6.1 Terra is structured at $1 per million input tokens, $5 for output tokens, and $0.05 for cached input tokens, making it an attractive alternative to its pricier counterparts. For comparison, GPT-6 Luna begins at $0.10 for input tokens, while flagship GPT-6 Astra is significantly more expensive, priced at $10 for input tokens.
Industry experts highlight that while cheaper processing improves economic elements, the overall cost-effectiveness of using models like Terra also depends on review and integration expenses. Ultimately, it comes down to finding a balance between quality output and cost.
