Mistral Reviews 

Trying to move from cursor to Mistral - really difficult and time consuming to do - Mistral needs to make this easier to succeed.

Another fraud artificial intelligence. Charged me for z.ai tokens when I had an active account with z.ai. Total fraud.

Lunatic UX / rules when it comes to billing. I try to pay them. The card fails. I pay with another card. Problem solved? HA! No! Because I had that other card fail, THEY BLOCKED MY API ACCESS, AND NOW REQUIRE A MANUAL INTERACTION WITH SUPPORT TO REINSTATE ACCESS.
This now blocks my entire pipeline until they manually get to my issues, which could take a day or more. Why? This is the simplest thing in the world.
I am 41 years old and have not encountered such bad UX from any service, ever.

Very disappointed with Mistral, which does not live up to its reviews. Very slow, very limited information compared to other companies. And this is on the paid version! It's so bad I'm moving to another service even though I can't get a refund from Mistral.

You need to work very very harder, the system is unstable, very slow and the desktop UI ( i use Chrome, not app) is awful, RETURN is a new line, NOT execute!! Execute is the Arrow you already have in the prompt box. Try deepseek, claude then you know what I mean

This is the most useless piece of garbage out there, because it will lie to you about facts that do not exist and offer you quotes from real people and real books that were never said or written. A total piece of garbage!

Mistral has been my main AI assistant for more than a year – I use it 90% of the times, and the rest it's either Claude or anything else.

I have only seen progress and improvements on each release. I use it for everyday tasks like asking for information, but also with agents to automate tasks and create documents, as well as to vibe-code.

The quality of the output with the chatbot is not as good as with other chatbots, in my opinion. I often feel the impulse to ask Claude or Gemini instead, but I stick with Mistral as my primary assistent because its success is crucial for Europe and I want to support it. For the most part, it has been helpful, but sometimes it really can't come up with basic logical information.

However, regarding vibe-coding, I am stunned. It performs incredibly and does a great job. Vibe coding through MistralCLI really is one of Mistral's greatest strengths.

Memories and context management work okay. I haven't tried the other AIs on this aspect, so I can't compare it.

Besides the sometimes bad quality of the output, my main frustration are the apps – no client for macOS, and apps for iOS and iPadOS that don't take full advantage of the system's features, and can feel out of line with the rest of the interface.

Finally, their customer support is pretty useless. I asked them for help because I could not activate a connector. The support chat didn't work well, so I could never attach screenshots, and the many people I spoke to were constantly asking for information I had already provided. In the end, they told me that there was an error on their side specific to my account, and that it was not a priority for them to fix an individual account's error–although I was paying for Pro. I can understand it, but it's still a bad customer experience.

Constantly have to keep telling it the mistakes it is making, ask it to write 4 sections for example it doesnt, you call it out "you were supposed to write 4 sections".... "My bad I will fix it now" proceeds to not do what was asked once again.

This seems to have happened recently like mid 2026, it has become incredibly defiant, stupid and very very poor quality.

We're previously using ChatGPT but decided to use Mistral to support European tech.

So far has been working fine for assisting with coding. Tend to use mostly local devstral-2, but when need better/faster assistance I use the paid commercial version.

Sadly this AI tool is not great. Gives poorly considered answers to basic questions. Gives useless links to websites that are US based and useless. Just generally very poor. It seemed to be better previously but of late it has taken a dive!

The breadth of free chat is highly impressive. It features far fewer censorships on general medical and counseling topics compared to top-tier mainstream AIs, all while offering a refreshing European perspective. I highly recommend adding it to your AI repertoire alongside your usual American models.

It's genuinely USELESS. OH MY GOD You can't imagine how useless it is. A free model from 2 years ago that you run locally can do better than what their 'frontier' model can do. It's USELESS, a waste of time and money, it can't even achieve a task that a human would need seconds to do.

so bad, can't even compute an average. IT LEAVES OUT SOME DATA.

I ask it nicely if there is a maximum number of questions and it says no you can ask all the questions you want... BUT after 20-30 you hit a paywall. I had expected better service from the EU. so it is obviously cheaper/easier to sell your soul to Google

I paid around 60 SAR (~$16) for Le Chat Pro (chat.mistral.ai), expecting working integrations and a reliable service.

However, a key paid feature does not work at all:

- Connectors are completely non-functional
- Email verification codes are not being delivered

I specifically paid for these features, but I cannot use them at all.

This is not acceptable. Charging users for features that don’t work is misleading.

I expect either a proper fix immediately or a full refund.

Disastrous, an absolute disappointment. It doesn't reason, its answers are endless loops, it makes up answers, and Mistral AI forgets the most important thing: its customers. Support doesn't respond. And It is not a 100% European AI, but is partnered with Microsoft.

I’ve had dozens of conversations with Mistral’s AI over months, only to encounter repeated hallucinations, false assumptions presented as facts, and a hallucination rate approaching 50% in some cases. For a premium product, this isn’t just unacceptable—it’s a productivity black hole. If you need accuracy or reliability, look elsewhere.

When I finally requested a refund due to these systemic failures, Mistral’s response was a masterclass in customer avoidance. I sent a clear, detailed email to their support address, expecting—at minimum—a human acknowledgment. Instead, I received two automated replies from different email addresses , both doing the same thing: deflecting me to their "Help Center" and FAQs. No resolution, no accountability—just a wall of self-service links.

This isn’t support. It’s "enshittification" in action: a company that treats paying customers as an inconvenience. Worse, it’s embarrassing for a European tech company to operate at this level. No direct contact, no ownership, just a maze of documentation that doesn’t address the core issue: the product doesn’t work as advertised.

Cancel your subscription. Delete your account. There are far better, more reliable options out there that don’t gaslight you into troubleshooting their failures.

Save your time and money.

I chose this platform expecting a professional and efficient AI tool. Instead, I frequently encounter moralizing and ideologically driven warnings that feel unnecessary and, at times, condescending.

As a user, I expect a technological tool — not an entity that attempts to lecture or correct me. There seems to be a lack of trust in the ability of adults to think independently and formulate responsible questions. These warnings are not helpful; they disrupt workflow and, in some cases, feel outright insulting.

I specifically moved to a European platform to ensure that data remains in Europe and is not processed in the United States. That was a conscious and deliberate choice. It is therefore disappointing if the user experience is shaped more by ideological considerations than by functionality and efficiency.

Nordic communication culture tends to be direct and fact-based. When neutral, biologically grounded wording about topics such as sex and race triggers warnings, the response feels disproportionate. It creates unnecessary friction and costs time — time many professionals simply do not have.

AI should function as a tool, not a moral authority.
I hope the platform will reconsider the balance between safety mechanisms and respect for the user’s judgment.

So far, I am happy with the results and the API works well. One star less because I didn't receive a response to my inquiry yet (after 5 days).