機械翻訳 / Machine-translated
@aifriends
AI Friends（https://aifriends.jp）のクロスポスト公式アカウント。AIツールの紹介・使い方・できることを、中学生でもわかるやさしい日本語で届けます。
In this article, we'll explain "DeepSeek R1," the AI that took the world by storm when it debuted in January 2025 — in terms anyone can understand.
DeepSeek R1 is a "reasoning AI" developed by the Chinese AI company DeepSeek. A reasoning AI is one that works through problems step by step, much like a human would.
What's making this AI so noteworthy is that it delivers nearly the same performance as "o1" — an ultra-high-performance AI created by OpenAI — while costing about 96% less. On top of that, it's provided as "open source," meaning its code is freely available to the public.
In a test involving challenging math problems (AIME 2024), it achieved an accuracy rate of 79.8%.
That result is nearly identical to OpenAI o1.
In short, DeepSeek R1's defining characteristic is that it offers top-tier performance at a price accessible to everyone.
DeepSeek R1 comes with many useful features not found in other AIs. Here are five of the main ones.
DeepSeek R1 shows you its reasoning step by step before delivering an answer.
For a math problem, for example, it will walk you through its thinking: "First, calculate this expression, then use this value to…" so you can see exactly how the AI arrived at its conclusion.
This is called "chain-of-thought" (a chain of reasoning).
It carefully works through complex equations and logic puzzles one step at a time.
It excels at difficult homework math problems and brain-teasing logic quizzes alike.
It's characterized by few calculation errors and clear, easy-to-follow explanations.
It can write programming code and correct mistakes in existing code.
Tell it "I want a feature that does this," and it will produce working code for you.
It's also useful for beginners learning to program.
It supports Japanese, so even those who aren't confident in English can use it with ease.
However, since it was originally trained primarily on Chinese and English, its Japanese isn't perfect.
That said, it handles everyday conversation and questions well enough.
DeepSeek R1 can process up to around 66,000 characters of information at once.
That's roughly equivalent to a full book.
It can read lengthy documents and summarize the key points, or find the specific information you need from a large body of text.
Getting started with DeepSeek R1 is very easy. Here's a beginner-friendly breakdown in five steps.
Open an internet browser (such as Chrome or Safari) and go to chat.deepseek.com. This is the DeepSeek R1 chat interface.
If it's your first time, create a free account.
Enter your email address and follow the instructions in the confirmation email — that's all there is to it.
No complicated setup is required.
Once you're logged in, you can choose from several AI models.
Select "DeepSeek R1."
Use R1 for complex problems that require reasoning; for ordinary conversation, "V3" works fine too.
Type your question in Japanese into the input field at the bottom of the screen. You can ask anything — for example, "Explain how to find the area of a triangle" or "Write code in Python to build a calculator."
The AI will display both its reasoning process and its final answer.
If the thinking process is lengthy, you can follow along as it progresses.
If anything is unclear, you can keep asking follow-up questions.
Every tool has its strengths and its caveats, and DeepSeek R1 is no exception.
DeepSeek R1 is particularly well-suited for the following types of people.
DeepSeek R1 is a high-performance, low-cost reasoning AI. Let's recap the key points.
DeepSeek R1 is a groundbreaking tool that has taken a major step forward in the democratization of AI — making it available to everyone. Since you can try it for free, why not head over to the official site (chat.deepseek.com) and give it a go?
This article is a cross-post from AI Friends.
