←back to thread

Devstral

(mistral.ai)
701 points mfiguiere | 8 comments | | HN request time: 0.001s | source | bottom
1. bravura ◴[] No.44054167[source]
And how do the results compare to hosted LLMs like Claude 3.7?
replies(1): >>44054278 #
2. resource_waste ◴[] No.44054278[source]
Eh, different usecase entirely. I don't really compare these.
replies(2): >>44055782 #>>44056372 #
3. bufferoverflow ◴[] No.44055782[source]
Different class. Same exact use case.
4. ttoinou ◴[] No.44056372[source]
For which kind of coding would you use a subpar LLM ?
replies(3): >>44056796 #>>44058368 #>>44059135 #
5. troyvit ◴[] No.44056796{3}[source]
I'd use a "subpar" LLM for any coding practice where I want to do the bulk of the thinking and where I care about how much coal I'm burning.

It's kind-of like asking, for which kind of road-trip would you use a Corolla hatchback instead of a Jeep Grand Wagoneer? For me the answer would be "almost all of them", but for others that might not be the case.

replies(1): >>44057291 #
6. ttoinou ◴[] No.44057291{4}[source]
In that case examples of which trips would be interesting so we can take inspiration from you
7. __MatrixMan__ ◴[] No.44058368{3}[source]
The kind of coding that happens after the internet goes down. How important of a use case that is depends heavily on why the internet went down.
8. kergonath ◴[] No.44059135{3}[source]
A LLM that I don’t host is a non-starter, so even a “subpar LLM” is better than someone’s cloud.