Can You Run Mistral Small 3.x 24B? VRAM Requirements
Mistral Small 3.x 24B has 24B total parameters. Running it at Q4_K_M (GGUF) with 16K context takes roughly 17.6 GB of memory (13.6 GB weights + 2.5 GB KV cache + runtime overhead). The minimum viable hardware is about RTX 3090 / 4090 24GB. All numbers are planning estimates, verified 2026-07-27.
VRAM by quantization (16K context)
| Quantization | Weights | Total needed | Minimum hardware |
|---|---|---|---|
| FP16 / BF16 | 44.7 GB | 50.8 GB | Mac 64GB unified |
| FP8 | 22.4 GB | 26.6 GB | RTX 5090 32GB |
| Q8_0 (GGUF) | 23.7 GB | 28.1 GB | RTX 5090 32GB |
| Q6_K (GGUF) | 18.3 GB | 22.3 GB | RTX 5090 32GB |
| Q5_K_M (GGUF) | 15.9 GB | 19.9 GB | RTX 3090 / 4090 24GB |
| MXFP4 | 11.8 GB | 15.8 GB | RTX 3090 / 4090 24GB |
| Q4_K_M (GGUF)recommended | 13.6 GB | 17.6 GB | RTX 3090 / 4090 24GB |
| Q3_K_M (GGUF) | 10.5 GB | 14.5 GB | RTX 4060 Ti 16GB |
| Q2_K (GGUF) | 7.8 GB | 11.8 GB | RTX 4060 Ti 16GB |
How context length changes memory (Q4_K_M (GGUF))
| Context | KV cache | Total needed |
|---|---|---|
| 4K | 0.6 GB | 15.8 GB |
| 16K | 2.5 GB | 17.6 GB |
| 64K | 10.0 GB | 25.1 GB |
| 128K | 20.0 GB | 35.1 GB |
KV cache grows linearly with context; FP8 KV cache halves it again (try it in the calculator below).
Customize the estimate
LLM VRAM Calculator — Can I Run It?
Estimate how much GPU memory (VRAM) you need to run open-weight LLMs like Kimi K3, DeepSeek R1, Qwen3, or Llama locally. Pick a model, quantization, and context length — the calculator adds up weights, KV cache, and runtime overhead, then shows which GPUs or Macs can fit it. Runs entirely in your browser: no upload, no signup.
24B total params · max 128K context
Recommended balance · ~0.61 bytes/param
Estimated memory needed
35.1 GB
Will it fit?
Assumes ~92% of device memory is usable. Multi-GPU counts are for tensor/pipeline parallel serving (vLLM, SGLang); Apple Silicon uses unified memory via llama.cpp or MLX.
FAQ
How much VRAM does Mistral Small 3.x 24B need?
About 17.6 GB at the recommended Q4_K_M (GGUF) quantization with 16K context: 13.6 GB for weights, 2.5 GB for KV cache, plus runtime overhead. Longer context grows the KV cache.
What is the minimum hardware for Mistral Small 3.x 24B?
Roughly RTX 3090 / 4090 24GB, assuming 92% of device memory is usable. Lower-end setups can try more aggressive quantization (Q3/Q2) at a noticeable quality cost.
Can I run Mistral Small 3.x 24B on a Mac?
Yes. At Q4_K_M (GGUF) with 16K context it needs about 17.6 GB, which fits the unified memory of a Mac M3 Ultra 512GB via llama.cpp or MLX.
Which quantization should I use for Mistral Small 3.x 24B?
Q4_K_M is the recommended balance of size and quality; with headroom, Q6_K or Q8_0 reduce quality loss further.