TL;DR
- 16GB: 7B models only (tight)
- 36GB: 13B comfortably, 32B Q4 tight
- 64GB: 32B Q5 comfortably
- 128GB: 70B Q5 comfortably
- Cannot upgrade after purchase — buy maximum at purchase time
How Unified Memory Works for LLMs
Unified memory is shared between CPU and GPU — all of it shared between the model and macOS. Unlike discrete GPU (RTX 4070 has 12GB VRAM + 32GB RAM separate), Apple Silicon shares one pool. 64GB Mac = 64GB in total. Plan at least 8GB for macOS and apps, leaving ~56GB for the LLM.
Is Unified Memory the Same as VRAM?
Yes — on Apple Silicon, unified memory *is* the GPU's VRAM. There is no separate, fixed VRAM pool like on a discrete GPU. The CPU, GPU, and Neural Engine on an M-series chip read and write the same physical memory, so nearly the entire unified memory pool — not a fraction of it — is available to the GPU for running an LLM, once at least 8GB is set aside for macOS. macOS also applies a default GPU memory limit, roughly two-thirds to three-quarters of total RAM depending on size (it can be raised), so treat total minus 8GB as an upper bound, not a guarantee.
- Discrete GPU (RTX 4070): 12GB VRAM is a hard ceiling for model weights, even if the PC has 64GB of separate system RAM.
- Apple Silicon: the GPU can address the full unified memory pool minus at least 8GB for macOS — a 64GB Mac gives the GPU ~56GB, not a fixed 12–24GB VRAM slice.
- This is why a 64GB Mac can run a 32B model comfortably while a 64GB PC with a 12GB GPU cannot — the PC's GPU-accessible memory is capped at its VRAM, not its total RAM.
- "Shared memory" and "unified memory" describe the same architecture: one pool, addressable by CPU and GPU alike, with no manual VRAM allocation step.
The Master Table: Memory Tier vs Model Size
Model | Parameters | Q3_K | Q4_K_M | Q5_K_M | Q8 | FP16 |
|---|---|---|---|---|---|---|
| Phi-4 Mini | 3.8B | 2.1 GB | 2.5 GB | 2.9 GB | 4.0 GB | 7.6 GB |
| Mistral 7B | 7B | 3.8 GB | 4.5 GB | 5.2 GB | 7.5 GB | 14 GB |
| Llama 3.1 8B | 8B | 4.2 GB | 5.0 GB | 5.8 GB | 8.5 GB | 16 GB |
| Gemma 3 12B | 12B | 6.5 GB | 7.8 GB | 9.0 GB | 12.9 GB | 24 GB |
| Qwen3 32B | 32B | 16 GB | 20 GB | 23 GB | 35 GB | 66 GB |
| Llama 3.3 70B | 70B | 36 GB | 42 GB | 49 GB | 74 GB | 140 GB |
| Llama 3.1 405B | 405B | 200+ GB | 240 GB | 280 GB | 410 GB | 810 GB |
Always subtract at least 8 GB for macOS when calculating fit on your Mac.
Fits / Doesn't Fit Matrix
Model + Quant | 16GB | 36GB | 64GB | 128GB |
|---|---|---|---|---|
| Phi-4 Mini Q4 (2.5 GB) | ✓ Plenty | ✓ Plenty | ✓ Plenty | ✓ Plenty |
| Llama 3.1 8B Q4 (5 GB) | ⚠️ Tight | ✓ Comfortable | ✓ Plenty | ✓ Plenty |
| Llama 3.1 8B Q8 (8.5 GB) | ✗ Won't fit | ✓ Comfortable | ✓ Plenty | ✓ Plenty |
| Gemma 3 12B Q4 (7.8 GB) | ✗ Won't fit | ✓ Comfortable | ✓ Plenty | ✓ Plenty |
| Qwen3 32B Q4 (20 GB) | ✗ Won't fit | ⚠️ Tight | ✓ Comfortable | ✓ Plenty |
| Qwen3 32B Q5 (23 GB) | ✗ Won't fit | ✗ Won't fit | ✓ Comfortable | ✓ Plenty |
| Llama 3.3 70B Q3 (36 GB) | ✗ Won't fit | ✗ Won't fit | ⚠️ Tight | ✓ Comfortable |
| Llama 3.3 70B Q4 (42 GB) | ✗ Won't fit | ✗ Won't fit | ⚠️ Very tight | ✓ Comfortable |
| Llama 3.3 70B Q5 (49 GB) | ✗ Won't fit | ✗ Won't fit | ✗ Won't fit | ✓ Comfortable |
| Llama 3.3 70B Q8 (74 GB) | ✗ Won't fit | ✗ Won't fit | ✗ Won't fit | ✓ Fits |
✓ Plenty = 4+ GB free | ✓ Comfortable = 2-4 GB free | ⚠️ Tight = under 2 GB free | ✗ Won't fit = uses swap or crashes
What Fits in Each Memory Tier (Practical)
- 116 GB (M5 base, MacBook Air)
Why it matters: Llama 3.1 8B Q4 fits (5GB model + 8GB OS = 13GB) ✓ but tight. Llama 8B Q8 won't fit without swapping. Whisper small fits alongside. - 236 GB (M5 Pro base)
Why it matters: Llama 3.1 8B Q8 fits comfortably. 13B-class models fit. Qwen3 32B Q4 barely fits (20GB + 8GB OS = 28GB). Multi-model: Whisper + LLaVA + TTS fit ✓ - 364 GB (M5 Pro max)
Why it matters: Qwen3 32B Q5 fits comfortably (23GB). Llama 70B Q3 barely fits. Multi-model stacks have plenty of room. - 4128 GB (M5 Max)
Why it matters: Best model for 128GB: Llama 3.3 70B Q5 (49GB) — the largest mainstream model that fits with room to spare. 70B Q8 fits (74GB) for near-lossless quality. Multi-modal: Whisper + 90B vision model + 8B LLM fit simultaneously ✓
Multi-Model Stack Memory Requirements
Stack Use Case | Memory Needed |
|---|---|
| LLM only (Llama 8B Q4) | 5 GB + OS = 13 GB |
| LLM + STT (Llama 8B + Whisper large-v3) | 8 GB + OS = 16 GB |
| LLM + STT + TTS (voice assistant) | 9 GB + OS = 17 GB |
| LLM + Vision (Llama 8B + LLaVA 7B) | 11 GB + OS = 19 GB |
| Full multimodal (LLM + Vision + STT + TTS) | 14 GB + OS = 22 GB |
| LLM + RAG (Llama 8B + embeddings + ChromaDB) | 8 GB + OS = 16 GB |
| Heavy multimodal (Llama 70B Q4 + Vision 90B) | 100+ GB |
Stacks above 22 GB need 36GB minimum Mac. Stacks above 50 GB need 64GB minimum Mac. The heavy multimodal stack only works on 128GB M5 Max.
Context Window Adds Memory Overhead
KV cache scales with context length — the longer the context window, the more memory your model uses at runtime. This is a common gotcha that can push a tight setup into swap.
- Llama 3.1 8B at 8K context: +0.5 GB
- Llama 3.1 8B at 32K context: +2 GB
- Llama 3.1 8B at 128K context: +8 GB
- Llama 3.3 70B at 32K context: +6 GB
- Llama 3.3 70B at 128K context: +24 GB
Buy Maximum Memory — Here's Why
- Cannot upgrade Apple Silicon memory after purchase.
- Model sizes are growing: 8B today → 13–32B sweet spot in 2027.
- 16GB is already marginal for LLMs — 36GB minimum recommended.
- Price difference: the 36GB→64GB upgrade costs extra at purchase (see Apple's current configurator for current prices) and may save buying a new Mac in 2 years if models outgrow 36GB.
- Example: M5 Pro 36GB vs 64GB — see Apple's current configurator for current prices; replacing the whole Mac later typically costs more than choosing the larger memory tier at purchase.
- A low-cost way to get 36GB+ unified memory: a Mac mini M5 Pro configured with 36GB has enough memory to run 13B models comfortably (see Apple's current configurator for pricing and configurations).
- Apple refreshed the Mac mini and Mac Studio lineup on August 25, 2026, adding tiers above this guide's 128GB ceiling: Mac mini M6 (32GB max, $899), Mac mini M5 Pro (64GB max, $1,699), Mac Studio M5 Max (128GB max, $2,499), and Mac Studio M5 Ultra (256GB or 512GB, from $5,499). The 512GB configuration ships late October 2026 above $10,000. For the model sizes this guide covers (up to 70B), 64–128GB is still the practical range — 256GB+ only matters for models larger than 70B or running several large models at once.
Quantization Impact on Quality
Q4_K_M (4-bit): quality loss is typically small and varies by model and task. A common default.
Q5_K_M (5-bit): quality loss is typically smaller than Q4 and varies by model and task. Recommended if you have spare memory.
Q8 (8-bit): quality loss is typically very small and varies by model and task.
Q3_K (3-bit): quality loss is typically larger and varies by model and task; it can be noticeable on complex reasoning. Acceptable only for space-constrained scenarios.
Should I get 36GB or 64GB?
Get 64GB if budget allows (see Apple's current configurator for the price difference). 36GB works today but may feel tight within 12 months as models grow. 64GB should cover the model sizes in this guide for the next year or two, if model sizes keep growing as they have.
Can I upgrade memory later?
No. Apple Silicon memory is soldered and non-upgradeable. Buy maximum at purchase time.
Why is 16GB not enough?
16GB total − at least 8GB for macOS = about 8GB available for the model. Llama 8B Q4 needs 5GB, leaving no room for Whisper or other tasks. Too tight.
Do I really need 128GB?
Only if you regularly run 70B models or need simultaneous vision + LLM + STT. Otherwise, 64GB is plenty.
Is 48GB enough for local LLMs?
Yes — 48GB (available on M4 Pro and some M5 Pro configs) is a comfortable middle ground. Runs all 32B models, 70B Q3 at the edge, and full multimodal stacks. Better than 36GB, but if you can afford 64GB, the future-proofing is worth it.
How much memory for running Llama 3.3 70B locally?
Minimum 48GB (Q3 quantization, noticeable quality loss). Recommended 64GB (Q4 quantization, tight fit). Comfortable 128GB (Q5/Q8 quantization, high quality). The 64GB tier requires careful memory management; 128GB is the only worry-free option for 70B.
Do I need 128GB for local AI in 2026?
Only if you're running 70B models regularly or need simultaneous vision + LLM + STT stacks. For everyday LLM use (8B-32B models, RAG, coding assistance), 64GB M5 Pro is the sweet spot. 128GB is a 2-3× price jump for marginal benefit unless you specifically need 70B.
Is Apple Silicon unified memory the same as VRAM?
Yes. Unlike a discrete GPU with a fixed, separate VRAM pool (e.g., 12GB on an RTX 4070), Apple Silicon's GPU reads and writes the same physical memory as the CPU. The GPU can access nearly all of a Mac's unified memory — a 64GB Mac gives the GPU ~56GB after the 8GB macOS reserve, not a fixed VRAM slice.
What's the largest LLM that fits in 128GB unified memory?
Llama 3.3 70B at Q5_K_M quantization (49GB) is the largest mainstream model that fits comfortably, leaving ~70GB free for context and overhead. At Q8 (74GB) it still fits with room for macOS. A 405B model needs 240GB+ even at Q4 — beyond a 128GB Mac's unified memory, so 70B is the practical ceiling on a 128GB Mac.
How much storage does a local LLM need, separate from memory?
Memory (RAM) and disk storage are different: memory holds the model while it runs; storage holds the downloaded model file permanently. A Q4 8B model needs ~5GB of both memory and disk. A 70B Q5 model needs ~49GB of memory and ~49GB of disk — the same file. Running several models locally (e.g., 8B, 32B, 70B, plus a vision model) needs 100-150GB of free disk space, independent of your Mac's memory tier.
Can I get more than 128GB of unified memory on a Mac?
Yes. Apple's August 25, 2026 refresh added the Mac Studio M5 Ultra with 256GB and 512GB unified memory configurations, starting at $5,499. The 512GB tier ships in late October 2026 and costs well above $10,000. That's far beyond what running 70B-class models requires — the 16GB–128GB tiers in this guide cover the practical range for nearly all local LLM use. 256GB+ only matters if you're running models larger than 70B or serving multiple large models simultaneously. No independent benchmarks exist yet for the M5 Ultra.