Can I run Mistral Small 3.2 24B locally?

Mistral Small 3.2 24B needs about 16.5 GB of graphics memory. It runs on 6 of the 9 graphics card setups on this page, such as the RTX 3090.

The cheapest graphics card on our list that runs it (it still needs the rest of a PC): GeForce RTX 3090 24GB, $1,370€1,224, used (median of sold listings).

Fast~35 words/sAnswers appear faster than you can read them.

Our price is for a used card. These shops mostly sell new ones, so their price can be higher.

The cheapest complete machine that runs it: Mac mini M6 32GB, $1,416€1,530. price unverified Slow~3.9 words/s

Which graphics cards run it

Seven popular graphics cards and two PCs with two cards each, at the standard quality (Q4). Click one for the full answer: memory, speed and price.

Runs well6

The whole model fits in the graphics memory.

Doesn't fit3

Not enough graphics memory: it needs about 16.5 GB. In a PC with enough normal memory it runs slowly: Budget PC with RTX 3060 12GB, $1,305€1,357. price unverified Slow~2.9 words/s

Every model on every card

Cheapest hardware for each size

From all 36 graphics cards, PCs and Macs on our list. "Runs slowly" means part of the model runs from normal memory: cheaper, but much slower. A smaller download needs less memory and answers a little less well.

  • Standard quality (Q4_K_M)GeForce RTX 3090 24GB, $1,370€1,224, used (median of sold listings)
  • Standard quality (Q4_K_M), runs slowlyBudget PC with RTX 3060 12GB, $1,305€1,357 price unverified
  • Smallest download (IQ2_M)GeForce RTX 3060 12GB, $288€257, used (median of sold listings)
  • Smallest download (IQ2_M), runs slowlyNo machine on our list runs this size.

Prices excl. shipping. Euro prices incl. VAT. Dollar prices are converted from our euro prices without EU VAT, before US sales tax.

Prices moved: see Building a local LLM rig in the RAMageddon.

About Mistral Small 3.2 24B

24.01B parameters. Parameter count is an estimate. More parameters usually means a smarter model that needs more memory.

Licence:
Apache-2.0 source
Longest conversation:
about 197 pages of text (131,072 tokens)
Input:
text + image in, text out
Technical notes

Mistral's mid-size open model for chat, documents and image input, under Apache 2.0. At our default quality it needs about 16.5 GB, so it just misses a 16 GB graphics card: a slightly smaller download fits one, and a 24 GB card holds the default.

Parameter source · Architecture source · Model page

Memory at each quality level

Each download of Mistral Small 3.2 24B is a different size. The total adds the file, the memory a conversation of about 12 pages of text (8,192 tokens) takes, and 0.8 GB of working space. A measured file size is labelled; the rest is our estimate. The last column is the cheapest machine that holds the total.

Swipe the table sideways to see every column, up to the cheapest machine.

Memory per quant for Mistral Small 3.2 24B
Quality (file)File sizeConversationWorking spaceTotalCheapest machine that holds it
BF1647.15 GBmeasured GGUF1.3 GB0.8 GB49.3 GBMac Studio M4 Max 128GB, $3,683€3,290 price unverified
Q8_025.05 GBmeasured GGUF1.3 GB0.8 GB27.2 GBMac mini M4 Pro 64GB, $2,116€1,890 price unverified
Q6_K19.35 GBmeasured GGUF1.3 GB0.8 GB21.5 GBGeForce RTX 3090 24GB, $1,370€1,224, used (median of sold listings)
Q5_K_M16.76 GBmeasured GGUF1.3 GB0.8 GB18.9 GBGeForce RTX 3090 24GB, $1,370€1,224, used (median of sold listings)
Q4_K_M14.33 GBmeasured GGUF1.3 GB0.8 GB16.5 GBGeForce RTX 3090 24GB, $1,370€1,224, used (median of sold listings)
IQ4_XS12.76 GBmeasured GGUF1.3 GB0.8 GB14.9 GBRadeon RX 9060 XT 16GB, $522€564, new (cheapest in-stock board on GPU PRIX's German board list)
Q3_K_M11.47 GBmeasured GGUF1.3 GB0.8 GB13.6 GBRadeon RX 9060 XT 16GB, $522€564, new (cheapest in-stock board on GPU PRIX's German board list)
IQ3_XXS9.28 GBmeasured GGUF1.3 GB0.8 GB11.4 GBRadeon RX 9060 XT 16GB, $522€564, new (cheapest in-stock board on GPU PRIX's German board list)
IQ2_M8.11 GBmeasured GGUF1.3 GB0.8 GB10.3 GBGeForce RTX 3060 12GB, $288€257, used (median of sold listings)

Waiting for prices to drop?

Get an email when GPU prices drop. No spam, unsubscribe anytime. Privacy

Speeds and memory are our estimate from the hardware specs, not a measured test. Prices are dated and may change. How we calculate