LocalAI Navigator

Hardware

From a phone to an always-on AI appliance, cheapest to most powerful. “Runs up to ~NB” is the largest model that fits in memory at Q4 compression. List refreshes automatically every week.

Budget

Windows laptop 16GB (integrated GPU)

Laptop
$650

No real GPU — runs on CPU + system RAM. Comfortable up to ~12B at Q4.

RAM 16 GBVRAM 2 GBusable ≈ 2 GBruns up to ~3B @ Q4

Android flagship (12GB)

Phone
$800

Runs 1–4B models via PocketPal / local runtimes. On-device, offline.

Unified 12 GBusable ≈ 8 GBruns up to ~13B @ Q4

iPhone Pro (8GB)

Phone
$1,000

1–4B models via LM Studio mobile / PocketPal. Neural engine helps.

Unified 8 GBusable ≈ 4 GBruns up to ~6B @ Q4

MacBook Air M4 (16GB)

Laptop
$1,099

Unified memory makes this punch above Windows peers. ~12B sweet spot.

Unified 16 GBusable ≈ 12 GBruns up to ~20B @ Q4

Mid-range

Desktop + RTX 4060 Ti (16GB VRAM)

GPU rig
$1,200

16GB VRAM fits ~24B at Q4 fast. Best price/performance discrete GPU.

RAM 32 GBVRAM 16 GBusable ≈ 16 GBruns up to ~26B @ Q4

Desktop + RTX 5080 (24GB VRAM)

GPU rig
$1,700

24GB VRAM runs ~32B at Q4 fast. Strong single-GPU coding rig.

RAM 32 GBVRAM 24 GBusable ≈ 24 GBruns up to ~40B @ Q4

MacBook Pro M5 (24GB)

Laptop
$1,799

Comfortable to ~24B. Great portable local-AI machine.

Unified 24 GBusable ≈ 18 GBruns up to ~30B @ Q4

Pro

Desktop + RTX 5090 (32GB VRAM)

GPU rig
$2,500

Fastest single consumer GPU. ~70B at Q4 with offload, ~32B fully on VRAM.

RAM 64 GBVRAM 32 GBusable ≈ 32 GBruns up to ~53B @ Q4

MacBook Pro M5 Max (64GB)

Laptop
$3,499

Runs ~70–110B at Q4 in unified memory. Portable workstation.

Unified 64 GBusable ≈ 48 GBruns up to ~80B @ Q4

Max

Nvidia DGX Spark (128GB unified)

AI appliance
$3,999

Desktop AI appliance built to stay on 24/7. 128GB unified memory, runs 70B+ and large MoE locally. The 'own your intelligence' endgame.

Unified 128 GBusable ≈ 96 GBruns up to ~160B @ Q4

Mac Studio M5 Max (128GB)

Desktop
$4,999

128GB unified memory holds 110B dense or large MoE at Q4. Quiet, low-power.

Unified 128 GBusable ≈ 96 GBruns up to ~160B @ Q4

Mac Studio M5 Ultra (256GB)

Desktop
$8,999

256GB unified — runs frontier-class MoE (Kimi K2.6, DeepSeek V4) at Q4. Top of the prosumer ladder.

Unified 256 GBusable ≈ 192 GBruns up to ~320B @ Q4