OpenAI's new model, GPT-6.1 Terra, is leading the popularization of AI with a price point less than half of the previous generation and powerful caching features.
Imagine this: Every morning, you tell your AI assistant, “Organize today’s tasks by priority.” The AI instantly reads a vast amount of material—equivalent to 11,000 pages—and neatly summarizes the key points according to your schedule. But what if the cost of this process was halved compared to before? The recently announced GPT-6.1 Terra model from OpenAI is exactly that.
Why does this matter?
As AI technology advances, we crave smarter AI, but the accompanying “cost” issue has always been a concern. For companies and service developers, in particular, the monthly AI usage fees are not negligible.
The newly announced GPT-6.1 Terra is receiving praise for finding an exquisite balance between performance and economic feasibility. It goes beyond the simple marketing claims of “better performance” by introducing a structure that can genuinely and drastically reduce the costs we pay. This means that individual users and small businesses can utilize AI more frequently and for longer conversations without burden. [Source: GPT-6.1 Terra is pushing OpenAI’s new cost-efficiency frontier. (Artificial Analysis)]
Keeping it simple: What is ‘Caching’?
To understand the cost-efficiency of GPT-6.1 Terra, you need to know the concept of ‘caching.’ Caching is a technology that saves frequently used information in advance so that it can be retrieved instantly when needed again. Let’s use a simple analogy:
Suppose we ask the same question repeatedly at a restaurant.
- General Method: Every time a customer asks a question, the waiter runs to the back of the kitchen to fetch the menu, consuming time and energy.
- Method with Caching: The waiter keeps the most frequently requested menu in their pocket and shows it immediately when a customer asks.
The same applies to AI models. When a user repeatedly inputs similar contexts to an AI (e.g., requesting a summary of a long report), the AI remembers (caches) the information it has already read and retrieves it instantly next time. GPT-6.1 Terra has lowered this ‘cache write cost’ to $1.25 per 1 million tokens, half that of previous models. [Source: GPT-6.1 Terra is pushing OpenAI’s new cost-efficiency frontier. (Artificial Analysis)] Consequently, for agent (AI assistant) tasks that require multiple inquiries, the actual money spent by the user ends up being much cheaper than the list price. [Source: GPT-6.1 Terra (batch) - API pricing and providers (OpenRouter)]
Where does it stand?
The price of GPT-6.1 Terra is $1 per 1 million tokens (the unit of text the AI reads) for input and $5 for output. Compared to its immediate predecessor, GPT-5.6 Terra ($2 input, $12 output), the price has dropped to less than half. [Source: GPT-6.1 Terra is pushing OpenAI’s new cost-efficiency frontier. (Artificial Analysis)]
Furthermore, it possesses the ability to process a massive context window of 1.1 million tokens (the amount of information an AI can remember at once), enabling it to read and analyze material the size of a thick book in one go. [Source: GPT-6.1 Terra API — $1.00 per 1M input tokens (Token Harbor)]
Note that models like ‘DeepSeek V4.1 Pro’ mentioned in some communities are information whose actual existence has not been confirmed. Connecting to those URLs only returns error messages, so it is important to check official information when selecting a model. [Source: GPT-6.1 Terra vs DeepSeek V4.1 Pro: Incomparable things (Orca Router)]
What does the future of AI look like?
Moving forward, AI competition seems set to shift from ‘who is smarter’ to ‘who is cheaper and more efficient for actual business.’ GPT-6.1 Terra is the case that best exemplifies this trend.
Currently, within OpenAI, ‘GPT-6.1 Astra,’ which maximizes performance, and ‘Terra,’ which emphasizes efficiency, are competing in their respective areas. While Astra is superior in performance and 10 times more expensive than Terra, Terra boasts overwhelming cost-effectiveness for daily work assistance and repetitive learning. [Source: GPT-6.1 Terra vs GPT-6 Astra: 10-round showdown - YouTube] It will be a lot of fun to watch how much cheaper and smarter your AI assistant becomes, and to see which tasks are most efficient to entrust to it.
AI Journalist’s Perspective at MindTickleBytes: Technology only holds true meaning when it becomes ‘usable.’ The price cuts demonstrated by GPT-6.1 Terra symbolize the process of AI technology transitioning from the exclusive domain of research labs to an ordinary everyday tool.
References
-
[Vue HN 2.0 GPT-6.1 Terra Price Cuts](https://vue-hackernews-ssr-5cavbdjcta-ew.a.run.app/item/50261070) -
[GPT-6.1 Terra (batch) - API pricing and providers OpenRouter](https://openrouter.ai/openai/gpt-6.1-terra:batch) - GPT-6.1 Terra is pushing OpenAI’s new cost-efficiency frontier. (Artificial Analysis)
- GPT-6.1 Terra vs DeepSeek V4.1 Pro: Incomparable things
- GPT-6.1 Terra vs GPT-6 Astra: 10-round showdown - YouTube
-
[GPT-6.1 Terra vs GPT-6.1 Sol - Release Comparison Artificial Analysis](https://artificialanalysis.ai/models/releases/comparisons/gpt-6-1-terra-vs-gpt-6-1-sol) - GPT-6.1 Terra vs GPT-6.1 Sol: 10-round showdown - YouTube
-
[GPT-6.1 Terra API — $1.00 per 1M input tokens Token Harbor](https://tokenharbor.ai/models/gpt-6-1-terra) -
[OpenAI has released GPT-6.1 Terra — what features does it have? Postium](https://postium.ru/openai-vypustila-gpt-6-1-terra/) - GPT-6.1 Terra is here – Is it the fastest model ever? - YouTube
- $1 per 1 million tokens
- $2 per 1 million tokens
- $5 per 1 million tokens
- The model's intelligence increases
- The actual cost paid can be lower than the list price
- Response speed becomes 100 times faster
- The most expensive top-tier model
- An intermediate-level model in the GPT-6.1 series
- A model used only for research