What Models?

Pick the right model for your GPU — in seconds.

Devstral Small 24B VRAM requirements: which GPUs can run it?

23.6B params 128K max context Tool use

The smallest GPU memory that runs Devstral Small 24B well is 16 GB (Q4_K_M, e.g. RTX PRO 2000 Blackwell). 72 of 132 GPUs run it well.

MMLU 72 SWE-bench Verified 46.8

VRAM needed per quantization

QuantizationWeights16K context32K contextFull 128K
Q4_K_M13.3 GB15.8 GB18.2 GB32.9 GB
Q8_023.3 GB25.8 GB28.2 GB42.9 GB

Weights plus fp16 KV cache. Leave ~1 GB headroom for the inference engine.

Run it locally

Each command downloads the GGUF from Hugging Face, then Ollama starts a chat and llama.cpp starts an OpenAI-compatible server with a web UI on port 8080.

Ollama ollama run hf.co/unsloth/Devstral-Small-2505-GGUF:Q4_K_M
llama.cpp llama-server -hf unsloth/Devstral-Small-2505-GGUF:Q4_K_M
Ollama ollama run hf.co/unsloth/Devstral-Small-2505-GGUF:Q8_0
llama.cpp llama-server -hf unsloth/Devstral-Small-2505-GGUF:Q8_0

Cheapest GPUs that can run Devstral Small 24B

Lowest launch MSRP among GPUs that run its smallest quantization (Q4_K_M) well, with at least 16K context.

As an Amazon Associate, I earn from qualifying purchases.

GPUs that can run Devstral Small 24B

Best quantization that fits on each GPU, with the context and speed it gets. For chips with several memory sizes, the smallest that works is shown.

NVIDIA

AMD

Intel

Apple

  • M1 16 GB Q4_K_M 17K context · ~4 tok/s
  • M1 Pro 16 GB Q4_K_M 17K context · ~13 tok/s
  • M2 16 GB Q4_K_M 17K context · ~7 tok/s
  • M2 Pro 16 GB Q4_K_M 17K context · ~13 tok/s
  • M3 16 GB Q4_K_M 17K context · ~7 tok/s
  • M4 16 GB Q4_K_M 17K context · ~8 tok/s
  • M5 16 GB Q4_K_M 17K context · ~10 tok/s
  • M6 16 GB Q4_K_M 17K context · ~10 tok/s
  • M3 Pro 18 GB Q4_K_M 30K context · ~10 tok/s
  • M4 Pro 24 GB Q4_K_M 69K context · ~18 tok/s
  • M5 Pro 24 GB Q4_K_M 69K context · ~20 tok/s
  • M1 Max 32 GB Q8_0 56K context · ~15 tok/s
  • M2 Max 32 GB Q8_0 56K context · ~15 tok/s
  • M3 Max 36 GB Q8_0 82K context · ~12 tok/s
  • M4 Max 36 GB Q8_0 82K context · ~16 tok/s
  • M5 Max 36 GB Q8_0 82K context · ~18 tok/s
  • M1 Ultra 64 GB Q8_0 128K context · ~21 tok/s
  • M2 Ultra 64 GB Q8_0 128K context · ~21 tok/s
  • M3 Ultra 96 GB Q8_0 128K context · ~22 tok/s
  • M5 Ultra 96 GB Q8_0 128K context · ~32 tok/s
Check your exact setup in the full tool →