Local AI

Local AI hardware guide — what you need to run LLMs locally

Not sure what hardware you need for local AI? This guide covers VRAM requirements by model size, CPU vs GPU trade-offs, Apple Silicon performance, RAM for context windows, storage for model files, and recommended builds for every budget.

VRAM requirements by model size

The most important specification for local LLM inference is VRAM (GPU memory). Here is how much you need for each model size at common quantization levels:

Model sizeQ4 (4.5 bit)Q8 (8 bit)FP16Recommended GPU
1-3B~2-3 GB~3-5 GB~4-7 GBAny GPU with 4+ GB
7-8B~5-7 GB~8-10 GB~14-16 GBRTX 3060 12 GB, RTX 4060
13-14B~8-10 GB~14-16 GB~26-28 GBRTX 3090 24 GB
30-34B~18-22 GB~32-36 GB~60-68 GBRTX 4090 24 GB, A6000 48 GB
70-72B~38-42 GB~68-74 GB~130-140 GB2x RTX 3090, A100 80 GB
120-180B~65-100 GB~115-175 GB~220-340 GB4x A6000, H100

Formula: VRAM ≈ (parameters × bytes per parameter) × 1.2 (overhead). At Q4: 4.5 bits = ~0.56 bytes per parameter. At Q8: 1 byte per parameter. At FP16: 2 bytes per parameter.

Context window overhead

In addition to model weights, you need VRAM for the KV cache. Each token of context consumes roughly:

  • 7B model: ~1 MB per 1K tokens of context
  • 13B model: ~2 MB per 1K tokens
  • 70B model: ~10 MB per 1K tokens

For an 8K context window with a 7B model, add ~8 MB extra VRAM. For 128K context with a 70B model, add ~1.3 GB.

CPU vs GPU inference

FactorCPU inferenceGPU inference
Speed2-8 tok/s (7B Q4)30-120 tok/s (7B Q4)
CostFree (existing hardware)$200-$3,000+ for GPU
Model size limitLimited by RAM (32-128 GB typical)Limited by VRAM (8-80 GB typical)
Energy efficiencyLower (CPU draws full power)Higher (GPU efficient for matrix ops)
Multi-userStruggles with concurrent requestsHandles batch inference well
Best enginellama.cpp, OllamavLLM, Ollama, llama.cpp (CUDA/Metal)

Recommendation: Use CPU inference for experimentation, small models (3B and under), or when you don't have a GPU. Use GPU inference for production-like workloads, larger models, and any use case requiring more than 10 tok/s.

Apple Silicon (M1, M2, M3, M4)

Apple Silicon Macs are excellent for local AI because of unified memory — the GPU and CPU share the same pool, so your model can access more than a typical GPU. llama.cpp with Metal acceleration achieves strong performance.

M-chipUnified memoryMax modelPerformance (7B Q4)
M1 (base)8 GB3B Q415-25 tok/s
M1 Pro16-32 GB8B Q420-35 tok/s
M1 Max32-64 GB13B Q425-40 tok/s
M2 Pro16-32 GB8B Q425-40 tok/s
M2 Max32-96 GB30B Q420-35 tok/s
M3 Pro18-36 GB13B Q430-50 tok/s
M3 Max36-128 GB70B Q315-25 tok/s
M4 Pro24-48 GB30B Q430-50 tok/s
M4 Max36-128 GB70B Q420-35 tok/s

Note: Apple Silicon + Metal via llama.cpp or Ollama is the most cost-effective way to run 30B+ models on consumer hardware. A Mac Studio with 128 GB unified memory can run 70B Q4 models that would require a $15,000+ multi-GPU workstation on PC.

RAM for context windows

System RAM matters for CPU inference and for storing context. When running on CPU or CPU+GPU hybrid (some layers on GPU, rest on CPU), the model weights are loaded into system RAM. Large context windows also consume system memory:

Context length7B model13B model70B model
4K (default)~16 GB RAM~32 GB RAM~80 GB RAM
8K~20 GB RAM~36 GB RAM~90 GB RAM
32K~32 GB RAM~48 GB RAM~130 GB RAM
128K~64 GB RAM~96 GB RAM~280 GB RAM

Rule of thumb: Double the model's VRAM requirement for system RAM. If a 7B Q4 needs 6 GB VRAM, ensure at least 16 GB system RAM. For a 70B Q4 needing 40 GB VRAM, ensure at least 64 GB system RAM.

Storage for model files

Model files are large. Plan your storage accordingly:

Model sizeQ4 file sizeQ8 file sizeFP16 file size
3B~1.8 GB~3.2 GB~6 GB
7-8B~4.5 GB~8 GB~15 GB
13-14B~8 GB~14 GB~27 GB
30-34B~18 GB~32 GB~62 GB
70-72B~40 GB~70 GB~140 GB

If you plan to experiment with 5-10 models, budget 100-200 GB of free storage. An NVMe SSD is strongly recommended — model loading time is bound by disk read speed.

Budget build (~$800) — entry-level local AI

ComponentRecommendedEstimated cost
GPUUsed RTX 3060 12 GB$200
CPURyzen 5 5600 / Intel i5-12400$120
RAM32 GB DDR4$60
Storage1 TB NVMe SSD$60
Rest (mobo, PSU, case)B550 / B660, 650W, ATX$360

Capable of: 7B-8B Q4 models at 30-50 tok/s, 13B Q4 at 15-20 tok/s (partial offload), 3B models comfortably. Good for learning, prompt engineering, and single-user agent prototypes.

Mid-range build (~$2,000) — serious local inference

ComponentRecommendedEstimated cost
GPUUsed RTX 3090 24 GB$700
CPURyzen 7 7800X3D / i7-13700K$350
RAM64 GB DDR5$180
Storage2 TB NVMe SSD$120
Rest (mobo, PSU, case)X670 / Z790, 850W, ATX$650

Capable of: 30B Q4 models at 15-25 tok/s, 70B Q3 at 8-12 tok/s, 7B-13B at Q8. Excellent for multi-model experimentation, local agents with RAG, and team of 2-3 users.

Workstation build ($5,000+) — production-like local setup

ComponentRecommendedEstimated cost
GPU2x Used RTX 3090 (48 GB total) or 1x A6000 48 GB$1,400-$4,000
CPURyzen 9 7950X / i9-14900K$550
RAM128 GB DDR5$350
Storage4 TB NVMe SSD$250
RestWorkstation mobo, 1200W PSU, large case$800

Capable of: 70B Q4 models at 15-25 tok/s, 120B Q3 at 8-12 tok/s, multiple concurrent models. Adequate for simulating production workloads, multi-user teams, and fine-tuning small models.

Note on cost

Before spending $2,000+ on hardware, consider that Plugsky's Enterprise plan starts at $249/month. At that price, it takes 8+ months of subscription to equal the workstation build — and you get 30+ models, 99.9% uptime, auto-scaling, and zero hardware maintenance. Use local hardware for prototyping and learning. Use Plugsky for production.

Not sure what hardware to buy?

Start with Plugsky's free tier — no hardware needed. Use it to test models and workloads before committing to a local hardware purchase.

Start Free → Back to Local AI →