Best Local AI Models by Apple Silicon Device

Browse AI model recommendations for every supported device: Apple Silicon Macs, iPhones, NVIDIA GPUs, AMD Radeon cards, and AMD Ryzen AI mini PCs.

Apple Silicon

Mac

MacBook Air
Apple M5·16 GB
M1M2M3M4M5
MacBook Pro
Apple M5 Pro·48 GB
M1M2M3M4M5
Mac Mini
Apple M6·16 GB
M1M2M4M6
Mac Studio
Apple M5 Max·64 GB
M1M2M3M4M5
On-device

iPhone

iPhone 15
Apple A16·6 GB
iPhone 15 Pro
Apple A17 Pro·8 GB
iPhone 15 Pro Max
Apple A17 Pro·8 GB
iPhone 16
Apple A18·8 GB
iPhone 16 Pro
Apple A18 Pro·8 GB
iPhone 16 Pro Max
Apple A18 Pro·8 GB
iPhone 16e
Apple A18·8 GB
iPhone 17
Apple A19·8 GB
iPhone 17 Air
Apple A19·12 GB
iPhone 17 Pro
Apple A19 Pro·12 GB
iPhone 17 Pro Max
Apple A19 Pro·12 GB
VRAM-bound

NVIDIA GPU

RTX 4060
8GB
MEMORY
8 GB GDDR6
SPEED
30 tok/s
ARCH
Ada Lovelace
BANDWIDTH
272 GB/s
Budget
RTX 3060
12GB
MEMORY
12 GB GDDR6
SPEED
42 tok/s
ARCH
Ampere
BANDWIDTH
360 GB/s
Budget
RTX 4060 Ti
16GB
MEMORY
16 GB GDDR6
SPEED
34 tok/s
ARCH
Ada Lovelace
BANDWIDTH
288 GB/s
Budget
RTX 5060 Ti
16GB
MEMORY
16 GB GDDR7
SPEED
51 tok/s
ARCH
Blackwell
BANDWIDTH
448 GB/s
Budget
RTX 4070
12GB
MEMORY
12 GB GDDR6X
SPEED
52 tok/s
ARCH
Ada Lovelace
BANDWIDTH
504 GB/s
Mid-Range
RTX 4070 SUPER
12GB
MEMORY
12 GB GDDR6X
SPEED
56 tok/s
ARCH
Ada Lovelace
BANDWIDTH
504 GB/s
Mid-Range
RTX 5070
12GB
MEMORY
12 GB GDDR7
SPEED
59 tok/s
ARCH
Blackwell
BANDWIDTH
672 GB/s
Mid-Range
RTX 5070 Ti
16GB
MEMORY
16 GB GDDR7
SPEED
87 tok/s
ARCH
Blackwell
BANDWIDTH
896 GB/s
Mid-Range
RTX 4070 Ti SUPER
16GB
MEMORY
16 GB GDDR6X
SPEED
72 tok/s
ARCH
Ada Lovelace
BANDWIDTH
672 GB/s
Mid-Range
RTX 4080 SUPER
16GB
MEMORY
16 GB GDDR6X
SPEED
79 tok/s
ARCH
Ada Lovelace
BANDWIDTH
736 GB/s
High-End
RTX 5080
16GB
MEMORY
16 GB GDDR7
SPEED
94 tok/s
ARCH
Blackwell
BANDWIDTH
960 GB/s
High-End
RTX 3090
24GB
MEMORY
24 GB GDDR6X
SPEED
87 tok/s
ARCH
Ampere
BANDWIDTH
936 GB/s
High-End
RTX 4090
24GB
MEMORY
24 GB GDDR6X
SPEED
104 tok/s
ARCH
Ada Lovelace
BANDWIDTH
1008 GB/s
High-End
RTX 5090
32GB
MEMORY
32 GB GDDR7
SPEED
145 tok/s
ARCH
Blackwell
BANDWIDTH
1792 GB/s
Ultra
RTX PRO 6000
96GB
MEMORY
96 GB GDDR7
SPEED
145 tok/s
ARCH
Blackwell
BANDWIDTH
1792 GB/s
Ultra
Radeon GPUs + Ryzen AI APUs

AMD

RX 7900 XTX
24GB
VRAM
24 GB GDDR6
MAX MODEL
32B parameter models
BANDWIDTH
960 GB/s
SPEED
89 tok/s
RDNA 3
RX 7900 XT
20GB
VRAM
20 GB GDDR6
MAX MODEL
27B parameter models
BANDWIDTH
800 GB/s
SPEED
74 tok/s
RDNA 3
Ryzen AI Max+ 395
110GB
MEMORY
110 GB unified
MAX MODEL
120B parameter models
BANDWIDTH
256 GB/s
SPEED
30 tok/s
Strix Halo
Getting Started Guides
Ollama Setup Guide

Install and run your first local AI model

Best LLM for MacBook

MacBook Air vs Pro, M1-M5 comparison

Best LLM for iPhone

Which AI models work on iPhone

Run AI Offline

Privacy-first AI without internet

Browse by Model

Qwen, Llama, DeepSeek, Mistral & more

Bar chart: maximum local LLM size by memory tier. 8 GB runs up to 9B, 12 GB runs up to 12B, 16 GB runs up to 14B, 24 GB runs up to 27B, 32 GB runs up to 35B, 36 GB runs up to 35B, 48 GB runs up to 35B, 64 GB runs up to 70B, 72 GB runs up to 70B, 96 GB runs up to 70B, 128 GB runs up to 70B, 192 GB runs up to 70B, 256 GB runs up to 70B, 512 GB runs up to 405B. Data from ModelFit's own catalog.

Not Sure Which Device You Have?

Use our interactive wizard to select your exact hardware configuration.

Open ModelFit Wizard