AI Creator's Path News: OpenAI o3 is now 80% off! AI coding has dramatically evolved, and now is the chance to improve development efficiency! #o3 #AIcoding #pricebreaking
Video explanation
A shock to the AI world! OpenAI's "o3" is now available at a shocking price, making your smart partner much more accessible?!
Hello, I'm John, a blog writer who loves the world of AI and technology! Recently, you may have heard more and more about "AI writing programs for you." It sounds like something from a science fiction movie, but it's already becoming a reality. Today, I'd like to tell you about some incredibly big news about an AI model called "o3" created by OpenAI, a company that has a reputation for being particularly "intelligent" among AI!
To put it in perspective, it's like suddenly being able to eat a full course meal at a super high-end restaurant that you can only go to on special occasions, for the price of the daily lunch special at your local diner...! This could completely change the world of AI-based manufacturing (especially programming)!
What is o3 and why is it so cheap?
First, let me briefly explain what o3 is. o3 is a kind of "intelligent AI assistant" developed by OpenAI that is good at understanding sentences, answering questions, and especially writing program blueprints (called code). It has been attracting attention for its intelligence, but it was a bit expensive.
However, just recently, OpenAI announced that the usage fee for o3 wasA whopping 80% discountAnd then they announced it! To be more specific, it's like a task that used to cost 1000 yen can now be done for just 200 yen. This is great news for programmers!
And it's not just the price that has gone down.Response speed (called latency, the time it takes for an AI to respond after you speak to it) has also improved significantly.It's like that. Previously, it was like, "Let's make a cup of coffee and wait...", but now it's running much faster. This is thanks to the introduction of a new super high-performance computer part (the latest GPU called GB200) made by a company called NVIDIA, and improvements to the way work is allocated to AI (called the scheduler).
How is it different from rival AI "Claude 4"?
There are many other smart AIs in the world of AI besides o3. For example, the AI "Claude 4" developed by Anthropic is also famous. Claude 4's appeal is that he responds quickly and can remember a lot of information at once (the technical term for this is "large context window").
However, some people who actually use it to write programs have said things like, "Claude 4 is fast, but sometimes he can be a little clumsy...He sometimes just creates the shell of a function that wasn't asked for, or rewrites test programs without permission."
On the other hand, the topic of this article is O3-kun, who is likeExperienced veteran craftsmenIt seems like. First, they think carefully, and sometimes they ask for confirmation, saying, "Is my understanding correct?", but in the end they carefully finish a program that works properly. Previously, people would say, "O3 is great at what it does, but the price is... like a supercar that's out of reach," but with this price revision, it has become so accessible that even people can afford it! It's like buying a used minicar.
Tips for using o3 wisely: Be careful of its little quirks!
Well, O3 is smart and affordable, but it seems to have some quirks.Tool Call" (a function that allows the AI to look up external information or run other programs on its own) seems to be used more than necessary. There are also times when users get frustrated, saying, "I wish it was simpler!"
But don't worry! Here are some tips to help you control your o3's "over-enthusiasm."
- Be specific with your requests and set limitsIf you clearly communicate the scope of what you want o8 to do and the number of times he can use a tool, such as "You may use the tool a maximum of eight times" or "Please only modify these two files," it will be easier for o2 to work without hesitation.
- Frequent checks and corrections：No matter how smart an AI is, it's not perfect. It's reassuring to know that the program that o3 created will be checked by a human and tweaked little by little as we go along.
- Don't forget to save frequently: This is the basic principle for any work, but if you save your work (called committing) frequently, if you ever think, "Huh? I think I went in a slightly different direction," you can quickly revert to the original state.
Does "thinking AI" really "think"?
AI like o3:InferenceIt is often said that AI is good at "reasoning". Reasoning is the ability to logically connect various pieces of information to arrive at an answer, just like humans carefully fitting together each piece of a puzzle. For example, this reasoning ability allows AI to perform complex tasks such as automatically correcting other related parts when you change one part of a program.
There is a term called "Chain-of-thought," which describes how AI solves problems by taking steps to arrive at an answer, just like humans do. It has been shown that using this method can dramatically increase the accuracy rate of AI programs.
However, there is an interesting story here. A research paper published by Apple points out that "AI doesn't really "think" like humans, but it may be something like a very intelligent "know-it-all parrot." In other words, it is just finding a pattern from the huge amount of data it has learned in the past that says "this is the most likely answer to this kind of question," and responding in that way.
Well, whether AI really "thinks" or not may be a philosophical question. But even if it's not "real thinking," it's an undeniable fact that AI's ability to help us with our work has improved dramatically! And the best part is that this amazing ability has become much more familiar to us with this price revision.
Amazing prices! What's behind them?
You might be wondering, "But how did they manage to make it so cheap? Is OpenAI okay?" In fact, there are a few big reasons behind this massive price cut, which could be called price destruction.
- Dramatic evolution of ultra-high performance computers：To run AI, you need an extremely powerful computer. The latest GPU (GPU, the part that acts as the brain of AI) called "GB200 NVL72" developed by a company called NVIDIA is said to be able to run AI 30 times more efficiently than previous models! What's more, it is surprising that it also uses less electricity.
- Greater cooperation between companies: A major factor is that a global technology company called Oracle has signed a huge contract to lend computer parts (chips) to OpenAI for over the next 10 years. This means that OpenAI no longer needs to make huge capital investments all at once, making it easier to keep costs down.
- Fierce competition to attract customers!: The world of AI development is currently in a period of war! Many companies are desperate to increase their customer base, saying, "We want as many developers and users as possible to use our AI while we can!" In a sense, they are offering attractive prices with a strategy of "Let's get people to use it first, even if it means a small loss!" It may be like a limited-time sale at a supermarket or a mobile phone company switching campaign.
Rivals aren't keeping quiet either! AI development is heading into an era of fierce competition
Of course, OpenAI isn't the only company at the forefront of AI development, with rival companies around the world eager to seize the opportunity.
- "BitNet b1.58" by Microsoft Research：This AI is so lightweight that you can write programs for it on a regular computer (meaning it runs on a CPU). It's a big step forward that we might no longer need expensive equipment to use AI.
- Alibaba's "Qwen3-235B-A22B": Alibaba, a giant Chinese IT company, is also developing a very smart AI. This model, called "Qwen," is said to have inference power comparable to that of Claude 4, and is designed to operate efficiently (using a technology called MoE: Mixture of Experts, which means combining multiple expert AIs).
With the emergence of new AIs one after another, AI technology will become cheaper and more powerful. In the future, using top-class AI will become commonplace, and the cost of using it will approach zero... that future may not be so far away!
How to deal with AI in the future: Tips for using o3 wisely
Now that the o3 is smart and affordable, we've put together some tips on how to use it in our work and creative endeavors.
- Let o3 be your main coder & planner!: The o3 used to be a bit slow and expensive, but that's no longer the case. The response speed is now tolerable, and the price is within reasonable limits. And above all, the intelligence of the software to think things through will ultimately produce high-quality results with fewer reworks.
- Leave the easy tasks to the more user-friendly AIIt may be a waste to ask o3 to do simple tasks like "Change the name of this file a little" or "Start Docker (a convenient box for running programs)". It is also a smart way to use a lighter and cheaper AI to quickly complete the task.
- Put a stop to o3's "too much tooling": As mentioned earlier, o3 has a tendency to overuse tools at times. You can control them by telling them rules like "Tools can only be used a maximum of xx times" at the beginning or by giving them detailed instructions.
- Be economical with prompts (instructions for AI)!: No matter how cheap it has become, it is a waste of time and money to have AI read unnecessary instructions or long sentences. Instead of telling it to "use the contents of this file as a reference," the trick is to keep instructions as concise as possible and focus on the main points, such as just telling it the file ID.
- Have a backup plan just in case: No matter how much AI advances, sometimes the server may be congested, slowing down response times, or unexpected errors may occur. In such cases, it is a good idea to have another AI model ready to use (for example, by setting it in your code editor).
- Try out various AI services：Famous AI coding support services such as Cursor and Windsurf are useful, but recently open source tools such as Roo Code and Cline have appeared that allow you to try out various AI models more freely. Trying to combine AI yourself might be fun as you make new discoveries!
Summary and some thoughts from John
I'm really amazed at how quickly AI technology is advancing. Until just a few weeks ago, I thought, "O3 is certainly smart, but it's a bit slow for everyday use, and the price is the worst..." However, with the price revision and performance improvement, my opinion has completely changed!
The response speed has also reached a practical level, making it easier on the wallet. And above all, its "thinking intelligence" may be more reliable than Claude 4, especially in terms of program reliability. With this, you can now use Gangan o3 as a reliable thinking assistant without having to worry about the contents of your wallet.
This OpenAI announcement is like an early Christmas present for the AI world. It's like Santa Claus, CEO Sam Altman (a big shot at OpenAI), has stuffed our stockings with "subsidized computing power (FLOPS, pronounced flops, a unit of computer computing speed)". Why not take this opportunity to ask the new o3 to do all sorts of "thinking" for you? I'm sure it will make your mental work a little easier!
This article is based on the following original articles and is summarized from the author's perspective:
OpenAI's o3 price plunge changes everything for vibe
coders