Mistral AI's 'Le Chat' Chatbot Now Available on iPhone

[MacRumors]

If you haven't had enough of AI apps, there's yet another to download and try out as of today. European company Mistral's Le Chat chatbot can now be used through a dedicated iOS app.

Mistral is a French AI company founded by engineers from Google and Meta. It creates its own open-weight large language models, and is aiming to compete with OpenAI. Le Chat has been available on the web, but the app will make it better able to compete with ChatGPT, DeepSeek, Gemini, and other options.

Like many competing chatbots, Le Chat supports natural language conversations, real-time web search, document analysis, and image generation.

LeChat is free to use, but access to the highest performing models is limited. A $14.99 per month fee unlocks a Pro tier with unlimited web browsing, extended access to news, and unlimited messaging.

Le Chat can be downloaded from the App Store for free. [Direct Link]

Article Link: Mistral AI's 'Le Chat' Chatbot Now Available on iPhone

 

[klasma]

I’m disappointed it doesn’t have a cat logo.

 

[CarletonTorpin]

Many cutting-edge LLMs actually don't use a 'mixture-of-experts' mode, yet, and Mistral does. This provides a differentiator for trying out Le Chat Ai, compared to other Ai's.

Apparently, it's tricky to implement the mixture-of-experts concept on existing LLMs, so they're needing to be rebuilt with that concept in mind, if those Ais want to harness the benefits it offers.

 

[dotnet]

I’m disappointed it doesn’t have a cat logo.

I wonder how many more "creatively" named AI assistants we will get. Chat Potato, anyone?

 

[It’s always something]

It’s also available for iPad.

 

[turbineseaplane]

Makes me think of this

 

[bsan]

Good company. I still have Mixtral 8x7B and Miqu on my desktop.
It's probably not objectively true, and Deepseek R1 has most of the spotlight now, but to me, they'll always be the original cool, open-source, and competitive AI lab.

 

[pinkkie]

Like many competing chatbots, Le Chat supports natural language conversations, real-time web search, document analysis, and image generation.

What is even the difference between all these? What is their selling point?

 

[AdamInKent]

Many cutting-edge LLMs actually don't use a 'mixture-of-experts' mode, yet, and Mistral does. This provides a differentiator for trying out Le Chat Ai, compared to other Ai's.

Apparently, it's tricky to implement the mixture-of-experts concept on existing LLMs, so they're needing to be rebuilt with that concept in mind, if those Ais want to harness the benefits it offers.

For those of us who don’t know, please could you explain the advantage of a “mixture of experts” mode?

 

[clementescolano]

I’m disappointed it doesn’t have a cat logo.

The orange logo seen in the screenshot is representing a cat, although in a stylised manner

 

[TheOldChevy]

What is even the difference between all these? What is their selling point?

They save their data in different countries and follow different legislations. As an example: https://arstechnica.com/security/20...-unencrypted-to-bytedance-controlled-servers/

Edit: add link.

 

[barfingclouds]

But I am le tired

 

[ApplesAreSweet&Sour]

What is even the difference between all these? What is their selling point?

One argument is that AI is very politicized now, and a lot of consumers are starting to alter their spending habits in relation to the trade wars:

Personally, I'm trying to lessen my contributions to the U.S. economy as I don't yet grasp how dedicating four years to make a single American and his businesses partners (even more) exorbitantly rich and powerful will help the country or its allies at large.

That's a level of 4D chess my woke, socialist brain cannot compute. Too many organic soy lattes served by DEI-hired trans baristas, I guess.

Anyway

-Not paying or generating data for OpenAI is a great selling point as they are one of Trump's partners.

Mistral also isn't that big of a step down. So it's not really much of a sacrifice.

 

[noannualfee]

I remember when they released Le Grill, the instructions were hard to follow.

 

[one more]

Have just tried it. The app typeface on an iPhone is a bit smaller compared to ChatGPT, but it can still come handy whenever ChatGPT complains that I have exceeded my daily request limit. Add Copilot and DeepSeek to the mix and everybody should be nicely covered for a day of free requests. 😉

 

[one more]

I’m disappointed it doesn’t have a cat logo.

The app icon looks kind of cattish, a-la a computer drawing from the 80s. 😉

 

[ArPe]

What is even the difference between all these? What is their selling point?

Some have bigger context lengths and bigger models are more reliable than smaller models. Some will waste your time showing how they come to an answer. They call this reasoning, but it is mostly a lot of babbling and sometimes the reasoning looks like a mental homeless person talking to an imaginary friend.

Anyway according to the AI clown show in the media before last month there were no open weight models people could use online or locally even though Ollama and Hugging Face have thousands to play with.

Those clown outlets said Deepseek “democratizes” AI but we always had models to download and the only way you can get close to ChatGPT level performance is if you have an extremely powerful computer or you use a model in the cloud. The biggest model you can run on a quad RTX 5090 still won’t be a ChatGPT, and if this is makes me look like a ChatGPT fan I’m not. I can live without all this jargon and hoopla.

 

[ArPe]

Have just tried it. The app typeface on an iPhone is a bit smaller compared to ChatGPT, but it can still come handy whenever ChatGPT complains that I have exceeded my daily request limit. Add Copilot and DeepSeek to the mix and everybody should be nicely covered for a day of free requests. 😉

you are fined one credit for violation of the verbal-morality statute​

 

[canadianreader]

Have just tried it. The app typeface on an iPhone is a bit smaller compared to ChatGPT, but it can still come handy whenever ChatGPT complains that I have exceeded my daily request limit. Add Copilot and DeepSeek to the mix and everybody should be nicely covered for a day of free requests. 😉

You could use Huggingface chat for free and load and use thousands of models including the ones from Mistral ai

HuggingChat

Making the community's best AI chat models available to everyone.

huggingface.co

 

[one more]

You could use Huggingface chat for free and load and use thousands of models including the ones from Mistral ai

HuggingChat

Making the community's best AI chat models available to everyone.

huggingface.co

Thank you, good to know!

 

[CarletonTorpin]

For those of us who don’t know, please could you explain the advantage of a “mixture of experts” mode?

To my understanding, a 'mixture-of-experts' model breaks a large Ai into several smaller, specialized parts.

When users ask a question, the system picks only the best-suited parts from the 'mixture of experts' and tasks them with answering the question.

The goal behind 'mixture-of-experts' models is to make the Ai's response faster and more accurate, compared to traditional methods for LLMs and Ais.

 

[CarletonTorpin]

Thank you, good to know!

Poe is great too, and has a free daily tier.

 

[CarletonTorpin]

you are fined one credit for violation of the verbal-morality statute​

So much for the three seashells.

 

[Aeronauts]

If it doesn’t know the answer does it shrug its shoulders and say “I do not know zee answer to this ridiculous question”

 

[Jarman74]

What is even the difference between all these? What is their selling point?

It might indeed be a good idea to recap what the main ones are doing, possibly around the time Apple Intelligence gets updated in April.