【Honest Review】The Real Truth About Using Mistral for One Month
機械翻訳 / Machine-translated
A
AI Friends
3日前・11分で読める
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。
What you'll learn in this article
I had been using ChatGPT Plus for about a year, but the $20 monthly fee was starting to feel like a burden. That's when I came across Mistral AI, a company from France. I heard that their Le Chat Pro plan was $14.99 per month, and that even the free plan let you use high-performance models, so I decided to give it a try. I also got a good impression from the fact that it's a company that values open source (a system that anyone can freely use). I was looking for an AI I could use for coding assistance and document creation while keeping costs down, so the timing was perfect.
The biggest surprise in the first week was how fast the responses were. Even when I asked Mistral the same questions as ChatGPT, Mistral felt about 1.5 times faster in returning answers. However, I did notice that there's still limited information available in Japanese. When I tried to look up how to use it, I had to read the official English documentation, which felt like a bit of a hurdle. Even so, the Le Chat chat interface is simple and easy to use, and I got used to it quickly. The coding assistance feature was also more impressive than I expected, with smooth code generation in Python and JavaScript.
1. Exceptional Value for Money
Le Chat Pro is $14.99 per month, while ChatGPT Plus is $20 and Claude Pro is similarly priced. When using the API (a system for accessing it from programs), the pricing is set at about 60% cheaper than other companies. Since even the free plan lets you use the high-performance Mistral Large model, it's also welcoming for those who want to try it out.
2. Response Speed Is Genuinely Fast
Even when asking technical questions or requesting summaries of long texts, I almost never had to wait. Since time is important when using it for work, this speed is a major advantage. In particular, the Mistral Medium 3.5 model released in April 2026 is an integrated model that combines chat, reasoning, and coding into one, making it possible to use without the hassle of switching between modes.
3. Active Addition of New Features
Since entering 2026, new features have been added one after another, including Voxtral for voice processing and a Vibe coding agent that automatically creates pull requests on GitHub. The enterprise-focused Mistral AI Workflows (a system for workflow automation) has also been released, and I feel optimistic about its future.
1. Limited Information Available in Japanese
When I search for help, there are still few explanatory articles or tutorials in Japanese. Since the official documentation is also primarily in English, people who are not comfortable with English may take some time to get started. While the AI's responses themselves support Japanese, I had difficulty looking up things like configuration methods.
2. Falls Behind Claude on Complex Reasoning Tasks
In situations requiring understanding of long contexts and solving complex problems, I felt Claude had higher accuracy. For example, in proofreading proposals that combined multiple conditions, or with intricate logical problems, Mistral would sometimes give answers that missed the point.
3. Thin Support for Enterprise Use
There's no problem using it personally, but the support structure and FAQ for companies looking to adopt it didn't feel fully developed yet. While adoption cases at large enterprises seem to be increasing, the contact point for small and medium-sized businesses to easily make inquiries was hard to find.
Compared to ChatGPT, my impression of Mistral is "fast and affordable." The quality of answers to the same questions is roughly equivalent, but the price is about 25% cheaper and the response speed is noticeably faster. On the other hand, compared to Claude, I felt Claude was superior for reading long texts and complex reasoning. However, when it comes to coding assistance, Mistral's Devstral model is extremely capable, with developer-focused features including GitHub integration well covered. In conclusion, it seems best to use Mistral for everyday use and coding, and Claude for analyzing difficult texts.
Recommended for
Not recommended for
This article is a cross-post from AI Friends.
