←back to thread

Devstral

(mistral.ai)
701 points mfiguiere | 1 comments | 21 May 25 14:21 UTC | HN request time: 0s | source
Show context
ManlyBread ◴[21 May 25 18:07 UTC] No.44054351[source]▶
>Devstral is light enough to run on a single RTX 4090 or a Mac with 32GB RAM, making it an ideal choice for local deployment and on-device use

This is still too much, a single 4090 costs $3k

replies(5): >>44054416 #>>44054504 #>>44054518 #>>44054679 #>>44062005 #
oezi ◴[21 May 25 18:22 UTC] No.44054518[source]▶
If it runs on 4090, it also runs on 3090 which are available used for 600 EUR.
replies(1): >>44054828 #
threeducks ◴[21 May 25 18:49 UTC] No.44054828[source]▶
More like 700 € if you are lucky. Prices are still not back down from the start of the AI boom.

I am hopeful that the prices will drop a bit more with Intel's recently announced Arc Pro B60 with 24GB VRAM, which unfortunately has only half the memory bandwidth of the RTX 3090.

Not sure why other hardware makers are so slow to catch up. Apple really was years ahead of the competition with the M1 Ultra with 800 GB/s memory bandwidth.

replies(1): >>44060286 #
1. paulbjensen ◴[22 May 25 09:14 UTC] No.44060286[source]▶
I managed to install and run it on my Razer Edge Laptop with an Nvidia RTX 4080, using Ollama.

It works but the tokens per sec is very slow. It did complete a TypeScript task example succinctly.