Ditched Claude and ChatGPT for Mistral
cross-posted from: https://lemmy.world/post/26431087
Since I canceled Claude (US boycott) I'm really enjoying https://chat.mistral.ai/ It might be a few months behind quality wise, but it is really fast, which is great for quick searches and brainstorming. It even has a canvas feature now.
This really gives me hope! Great french AI Tool!
Thanks. I use chatGPT a lot and I feel bad about it. I will check out Mistral.
I have to admit, France really stepped up in the past couple of years (and decades regarding nuclear deterrent ;) ). Besides Mistral.ai instead of US AI services, I started to use qwant.com as a replacement for Google search/news. It's been only a couple of days, but I'd say the quality is about the same/even better (but might be biased due that ongoing Trump shitshow).
The sad thing is that i am pretty fond of Claude. But i will give Mistral another go.
Mistral is the first ChatBot I started using since I did not trust the others from day one. Le Chat is pretty good IMHO for AI starters, although I do not have any comparison with other Bots (I don’t even know what „canvas generation“ means LOL).
The only downside is, that it’s currently only freely usable on an iPhone. There is a Mac App (Meow), but you will need a paid API Key to use it. If anyone knows a method how to use this free of change on a Mac, I would appreciate it. Thanks!
I'm avoiding AIs because of the power and water consumption. In my particular opinion they are not realiable and not worth the resources, yet. If at all I used them to generate logos and images, to avoid licensing troubles. Ah.. also sorting thorugh some data or producing structured data that follows a pattern.
Anyhow, a little comparison would be a good thing.
Tbh. U have been using perplexity as my daily driver to research stuff on the web.
For example, i am German and recently had to interact with our local bureaucracy on a topic i had no experience with. After hours of trying to figure out what to do, I asked perplexity. Not only did it find the right request form, it gave me detailed instructions on how to use it, what the field actually mean And Where to get the correct Values.
All in all using Perplexity properly saved me like 20 hours of trying to figure things out myself
@a887dcd7a @TheProtagonist I find it to be worth it for my usecases(code, analyses, websearch and as a french tutor) especially because their flash option does a lot for my UX. I found other options to be jarring at times. They think and think and the say aomething went wrong. Or worse do so halfway in providing an
answer.
I'm having high hopes for their upcoming reasoning model. If they can keep the flash option with that model(even in diminished capacity) it wi be mindblowing. But we will have to wait and see.
Y'all, what do you need an AI chatbot for anyway? European or not, those things have a huge power consumption, can't tell hallucinations apart from reality, and steal labor from common folk to make it available for free to rich people.
What's the use case? If you need an expert to help you out, you can talk to them, comission art from them, research yourself..
Y’all, what do you need an AI chatbot for anyway?
Developing little one off scripts that I wouldn't write otherwise. My home automation and the organisation of my media library has benefited greatly from that.
I can definitely NOT pay a developer for that, hourly rates start at 150 to 170€ with a few hours minimum.
Actually they are great for tedious tasks. For a CSV file I asked an AI to write down every hour of the day in steps of 15 minutes and provide a file to download.
It took only 5 attempts until all hours were included 😂
this can be accomplished in two seconds with an excel sheet
I use them to help me create incredibly niche Firefox addons that only I would ever care about (confirmed by the fact I've put them in the Firefox addons site and got no downloads at all in months/years) but are useful to me.
Yeah I could learn to make them myself. But why. I don't want to try to make big ones or make them regularly or anything. It would honestly be a waste of my time to learn. And I'm sure as hell not paying someone to make them.
Also they can be good at finding stuff. I was trying to buy something that was sold out everywhere and had searched like the first 3 pages of google results and I decided to ask if it could find one in stock and it found this super small site for this small store that had 2 left in stock.
So I asked it how it compares to ChatGPT, and got this response:
I am Le Chat, an AI assistant created by Mistral AI. While both ChatGPT and I are designed to assist users with a wide range of tasks, there are some differences in our functionalities:
Knowledge Cutoff: ChatGPT's knowledge is up-to-date as of 2021, whereas my knowledge base was last updated on October 1, 2023. This means I can provide more recent information.
Web Browsing: I have the ability to perform web searches to find up-to-date information, which allows me to provide real-time data. ChatGPT does not have this capability.
Multimodal Capabilities: I can read images, but I cannot generate images. ChatGPT does not have image reading capabilities.
Canvas Generation: Neither ChatGPT nor I have access to canvas generation mode.
Python Code Interpreter: Neither ChatGPT nor I can access the Python code interpreter.
Language Support: Both ChatGPT and I can respond in multiple languages, but I will only respond in English if I cannot infer the expected language from the user's message.
These differences highlight how my functionalities are designed to provide a more up-to-date and versatile user experience.
Not very exact. ChatGPT can perform web searches and read images.
This is exactly the type of answer I hate AI for. Outdated (Cutoff points) and wrong (image generation).
Nice query. Could you ask about the others (I don't now them all.. Gemini, Grok, ...). This could be a nice (even low key) overview post about differences.
How do you compare to Gemini, Claude and Grok AIs? Comparison of AI Models: Le Chat, Gemini, Claude, and Grok
Le Chat
Developer: Mistral AI
Capabilities:
Multilingual support Web browsing for up-to-date information Multi-modal interactions (text, images) Tool use for specific tasks Strong emphasis on safety and ethical guidelines
Gemini
Developer: Google DeepMind
Capabilities:
Multimodal capabilities, handling text, code, audio, image, and video
Optimized for various sizes: Ultra, Pro, and Nano
Advanced reasoning and tool use, including native image creation and speech generation
Integration with Google Search and other tools for enhanced functionality
AI-powered search features, such as AI Mode, for in-depth answers and complex queries
Claude
Developer: Anthropic
Capabilities:
Versatile problem-solving in coding, writing, and creative tasks
Strong ethical framework and transparency
Multimodal capabilities, accepting text, audio, and visual inputs
Specialized models for different use cases, such as Claude Code for coding tasks
High security with SOC 2 Type II certification and HIPAA compliance options
Grok
Developer: xAI (associated with Elon Musk)
Capabilities:
Advanced reasoning and extensive pretraining knowledge
Customizable response styles and personalized interactions
Uncensored voice modes for various interactions, including storytelling and therapy simulations
Integration with various platforms and continuous learning capabilities
Summary
Each AI model has its unique strengths and capabilities:
Le Chat excels in multilingual support and up-to-date information retrieval.
Gemini offers advanced multimodal capabilities and integration with Google services.
Claude focuses on ethical problem-solving and specialized models for various tasks.
Grok provides customizable interactions and advanced reasoning capabilities.
The choice between these models depends on the specific needs and preferences of the user.