With a discrete GPU, system RAM matters in two ways: model files load through it, and any layers that do not fit in VRAM run from it (slowly). Without a GPU, or on unified-memory machines, RAM holds the whole model.
Estimated memory by model
| Model | Params (total / active) | Est. memory, Q4 @ 8k ctx | Est. memory, Q8 @ 8k ctx | Cheapest fit in our data (Q4) |
|---|---|---|---|---|
| Llama 3.1 8B | 8.03B / 8.03B | 7.0 GB | 10.6 GB | GeForce RTX 5060 ($299, GPU launch MSRP) |
| Qwen3 14B | 14.8B / 14.8B | 11.5 GB | 18.2 GB | GeForce RTX 5060 Ti 16GB ($429, GPU launch MSRP) |
| Qwen3 32B | 32.8B / 32.8B | 22.5 GB | 37.3 GB | GeForce RTX 3090 ($1,499, GPU launch MSRP) |
| Qwen3 30B-A3B (MoE) | 30.5B / 3.3B | 21.1 GB | 34.9 GB | GeForce RTX 3090 ($1,499, GPU launch MSRP) |
| Llama 3.3 70B | 70.6B / 70.6B | 46.8 GB | 78.7 GB | GMKtec EVO-X2 (128GB / 1TB) ($3,499.99, system, listed price) |
| gpt-oss-120b (MoE) | 117B / 5.1B | 64.7 GB | 126.8 GB | GMKtec EVO-X2 (128GB / 1TB) ($3,499.99, system, listed price) |
Parameter counts: Hugging Face: openai/gpt-oss-120b model card; Hugging Face: meta-llama/Llama-3.1-8B; Hugging Face: meta-llama/Llama-3.3-70B-Instruct; Hugging Face: Qwen3 model cards (e.g. Qwen/Qwen3-30B-A3B). Memory figures are our estimates (see methodology), not measurements.
Do you need 128GB RAM?
128 GB matters when the RAM itself holds the model: on unified-memory systems (Ryzen AI Max+ 395, DGX Spark, Mac Studio) it allows 70B dense models at 4-bit or large mixture-of-experts models such as gpt-oss-120b (estimates above). On a desktop with a discrete GPU, 128 GB of system RAM mainly helps with CPU offload, which is much slower. Current 128GB kit prices: see RAM prices.
Rule of thumb: on a GPU system, have system RAM at least equal to the model file size; on CPU-only or unified memory, have the estimated total plus 8–16 GB for the OS and apps. Unified-memory systems:
| System | Chip | Memory | Memory bandwidth | Price | Source |
|---|---|---|---|---|---|
| GMKtec EVO-X2 (128GB / 1TB) | AMD Ryzen AI Max+ 395 | 128 GB unified | 256 GB/s (computed: 256-bit × 8000 MT/s) | $3,499.99 (maker store, accessed 21 Oct 2026) | GMKtec: EVO-X2 product page (128GB + 1TB variant); AMD: Ryzen AI Max+ 395 specifications |
| NVIDIA DGX Spark (128GB) | NVIDIA GB10 Grace Blackwell | 128 GB unified | 273 GB/s | $4,699 (NVIDIA Founders Edition MSRP from Feb 2026; current marketplace price TBD) A 64GB configuration is sold through OEM partners. | NVIDIA: DGX Spark product page and specs; NVIDIA Developer Forums: DGX Spark price change announcement (Feb 2026) |
| Mac mini (M5 Pro, 64GB) | Apple M5 Pro | 64 GB unified | 307 GB/s | TBD | Apple: Mac mini technical specifications |