Rent a GPU

Can the NVIDIA RTX 3090 run Magistral Small 24B?

RTX 3090 × Magistral Small 24B8K context
YesRuns fully on the GPU
Memory at Q6_K21.3 GB of 24 GB
Weights 19.4 GBKV cache 1.3 GBRuntime 0.6 GB
QuantQ6_KExcellent
Speed~29tokens per second
Max context23Ktokens that fit at Q6_K
  1. FP16
  2. Q8_0
  3. Q6_K
  4. Q5_K_M
  5. Q4_K_M
  6. Q3_K_M
  7. Q2_K

Yes. The NVIDIA RTX 3090 can run Magistral Small 24B fully in VRAM at Q6_K (excellent), using about 21.3 GB of the 24 GB available, at roughly 29 tokens/second.

Try other context lengths

Run it with Ollama

With Ollama installed, one command downloads and starts Magistral Small 24B:

ollama run magistral:24b

Ollama's default download is Q4_K_M (about 16.2 GB here). You have room for Q6_K, which is excellent: pick that GGUF in LM Studio or from Hugging Face.

New to this? Step-by-step setup for the RTX 3090

Run it in the cloud

No NVIDIA RTX 3090? Run Magistral Small 24B in the cloud.

Rent a cloud GPU by the hour. No upfront cost, stop any time.

Rent a GPU on RunPodRent a GPU on Vast.ai

Memory needed by quantization

At 8,192 tokens of context. The black line marks the 24 GB the NVIDIA RTX 3090 can use.

QuantQualityMemoryFits?Speed
FP16Full quality49.1 GBToo big—
Q8_0Near-lossless27 GBToo big—
Q6_KExcellent21.3 GBFits~29 tok/s
Q5_K_MVery good18.7 GBFits~33 tok/s
Q4_K_MGood, most popular16.2 GBFits~39 tok/s
Q3_K_MNoticeably degraded13.5 GBFits~49 tok/s
Q2_KHeavily degraded10.8 GBFits~63 tok/s

Measured speed from real RTX 3090 owners

Tokens per second that people measured on their own NVIDIA RTX 3090, next to our estimate.

No one has measured Magistral Small 24B on the NVIDIA RTX 3090 yet. The speeds above are estimates. Run the benchmark and yours becomes the first real number on this page.

Have an RTX 3090? Add your real speed

One command times Magistral Small 24B in your own Ollama, shows you exactly what it would send, and asks before sending anything. No sign-up, no personal data.

macOS and Linux

curl -fsSL https://gpufit.dev/bench.sh | sh -s -- magistral:24b

Windows (PowerShell)

$env:GPUFIT_MODEL="magistral:24b"; irm https://gpufit.dev/bench.ps1 | iex

How the benchmark works

GPUs that run Magistral Small 24B comfortably

Models that run well on the NVIDIA RTX 3090

Frequently asked questions

Can NVIDIA RTX 3090 run Magistral Small 24B?

Yes. The NVIDIA RTX 3090 can run Magistral Small 24B fully in VRAM at Q6_K (excellent), using about 21.3 GB of the 24 GB available, at roughly 29 tokens/second.

How much VRAM does Magistral Small 24B need?

About 16.2 GB at Q4_K_M with 8,192 tokens of context, and about 49.1 GB at full FP16 precision.

How fast is Magistral Small 24B on the NVIDIA RTX 3090?

Roughly 29 tokens per second at Q6_K, estimated from the NVIDIA RTX 3090's 936 GB/s memory bandwidth.