←back to thread

Le Chat: Custom MCP Connectors, Memories

(mistral.ai)
397 points Anon84 | 3 comments | 04 Sep 25 11:04 UTC | HN request time: 0s | source
Show context
Alifatisk ◴[04 Sep 25 12:30 UTC] No.45126512[source]▶
I use Qwen Chat as my daily, it’s good and does the job very well. I have never thought about trying out Mistral, is their model good at anything? Any area it excels at? Or is it far behind all the other models?
replies(2): >>45126743 #>>45128654 #
epolanski ◴[04 Sep 25 12:57 UTC] No.45126743[source]▶
Qwen is good for reasoning/feedback/creative work. I use it a lot when reviewing documentation, prose and some code, it's the one I like the most.

But when it comes to researching information is consistently among the worst performers in my comparisons.

Generally the order is Opus 4.1 > Perplexity > Gemini Pro >> GPT 5 >> Qwen.

replies(2): >>45126858 #>>45129746 #
Alifatisk ◴[04 Sep 25 17:22 UTC] No.45129746[source]▶
I got Perplexity Pro one year for free, what makes Opus 4.1 search so good? Is it worth switching?
replies(1): >>45130853 #
1. epolanski ◴[04 Sep 25 18:51 UTC] No.45130853{3}[source]▶
I don't know whether I can give you a conclusive answer, due to how query-dependent the results are and the lack of determinism.

I really like perplexity and if you get it free it's hard to justify spending 100$/month.

replies(1): >>45131436 #
2. Alifatisk ◴[04 Sep 25 19:43 UTC] No.45131436[source]▶
I understand, I thought maybe there was a couple of cases you stumbled upon that Opus 4.1 was better at, which made you rank it higher than Perplexity!
replies(1): >>45135881 #
3. epolanski ◴[05 Sep 25 07:26 UTC] No.45135881[source]▶
Yes this happened.

Few recent examples:

- find me public companies listed in Poland that have the best z-index of dividend yield, payout ratio and earnings growth

- list the most important psychological tendencies and biases that crew resource management tries to address