← Interactive LLM VRAM Calculator
Model VRAM guides
Approximate VRAM needed to run popular local LLMs at each GGUF quantization level (FP16, Q8_0, Q6_K, Q5_K_M, Q4_K_M) with 8k context, and example GPUs that fit. For custom context lengths and quantizations, use the interactive calculator.
- How much VRAM to run Llama 3.1 70B? (~47 GB at Q4_K_M)
- How much VRAM to run Llama 3.1 8B? (~7.0 GB at Q4_K_M)
- How much VRAM to run Llama 3.3 70B? (~47 GB at Q4_K_M)
- How much VRAM to run Qwen2.5 32B? (~23 GB at Q4_K_M)
- How much VRAM to run Qwen2.5 72B? (~51 GB at Q4_K_M)
- How much VRAM to run DeepSeek-R1 70B (Llama distill)? (~47 GB at Q4_K_M)
- How much VRAM to run DeepSeek-R1 671B? (~410 GB at Q4_K_M)
- How much VRAM to run Mistral 7B? (~6.5 GB at Q4_K_M)
- How much VRAM to run Mixtral 8x7B? (~28 GB at Q4_K_M)
- How much VRAM to run Gemma 2 27B? (~21 GB at Q4_K_M)
- How much VRAM to run Phi-4 14B? (~12 GB at Q4_K_M)
- How much VRAM to run Qwen2.5-Coder 32B? (~23 GB at Q4_K_M)