What AI models can Apple M3 Pro (18GB Unified) run?

The Apple M3 Pro (18GB Unified) can run 59 AI models. Top performers include all-MiniLM-L6-v2, nomic-embed-text v1.5, Gemma 4 26B-A4B. See the full compatibility table above for speeds and quality ratings.

Is Apple M3 Pro (18GB Unified) good for AI coding?

Yes. With 18 GB, the Apple M3 Pro (18GB Unified) handles single-model coding workflows well at the Capable tier.

How much VRAM does Apple M3 Pro (18GB Unified) have?

The Apple M3 Pro (18GB Unified) has 18 GB of unified memory with 150 GB/s bandwidth.

Can Apple M3 Pro (18GB Unified) run 70B models?

70B models can run on the Apple M3 Pro (18GB Unified) with CPU offloading, but performance will be reduced. Consider a GPU with 48GB+ VRAM for full-speed 70B inference.

Is Apple M3 Pro (18GB Unified) worth it for AI?

At $1,799, the Apple M3 Pro (18GB Unified) offers 18 GB VRAM and runs 59 AI models. It works for smaller models and experimentation.

Apple Silicon

Apple M3 Pro (18GB Unified)

18 GB Unified · 150 GB/s

From

$1,799

Estimated street price

VRAM

18 GB

Bandwidth

150 GB/s

TDP

30W

Models

Tier

Capable

The Apple M3 Pro (18GB Unified) with 18 GB unified memory can handle 59 AI models across embedding, ai_building, coding. Best performance: all-MiniLM-L6-v2 at 1200 tok/s (good). For AI coding workflows, it supports the Capable AI Coding tier, handling single model workflows well. Current price: approximately $1,799.

Source: OwnRig methodology

VRAM

18 GB

Bandwidth

150 GB/s

Memory Type

Unified

TDP

30W

GPU Cores

Host Devices

MacBook Pro 14", MacBook Pro 16"

Builder Capability: Capable AI Coding

Runs 16-22B coding models comfortably, or 32B at reduced quality. Handles single model workflows well.

Software

Inference Backends

The software stacks that matter most for real-world inference on this device.

Metal

production

Primary Apple Silicon backend across MLX and llama.cpp workloads.

What it can run

59 models


all-MiniLM-L6-v2	FP16	1200 tok/s	Good
Arcee Trinity Mini 26B	Q4_K_M	8 tok/s	Not viable
Arcee Trinity Nano 6B	Q8_0	32 tok/s	Good
Code Llama 34B Instruct	Q3_K_M	–	Not viable
Codestral 22B	Q3_K_M	–	Not viable
Command R 35B	Q3_K_M	–	Not viable
DeepSeek Coder V2 Lite 16B	Q4_K_M	20 tok/s	Good
DeepSeek R1 Distill Qwen 32B	Q3_K_M	–	Not viable
DeepSeek R1 Distill Qwen 7B	Q4_K_M	14 tok/s	Acceptable
DeepSeek V3	Q2_K	–	Not viable
FLUX.1 Dev	Q4_K_M	–	Not viable
Gemma 2 27B Instruct	Q4_K_M	–	Not viable
Gemma 2 9B Instruct	Q4_K_M	13 tok/s	Acceptable
Gemma 3 12B	Q3_K_M	5 tok/s	Marginal
Gemma 3 27B	Q3_K_M	3 tok/s	Marginal
Gemma 3 4B	Q4_K_M	22 tok/s	Acceptable
Gemma 4 26B-A4B	Q3_K_M	51 tok/s	Good
Gemma 4 31B	Q3_K_M	7 tok/s	Marginal
Gemma 4 E2B	Q8_0	22 tok/s	Acceptable
Gemma 4 E4B	Q8_0	14 tok/s	Marginal
GigaChat Lightning 10B	Q8_0	38 tok/s	Acceptable
InternLM 2.5 7B Chat	Q4_K_M	15 tok/s	Acceptable
Llama 3.1 70B Instruct	Q2_K	–	Not viable
Llama 3.1 8B Instruct	Q4_K_M	15 tok/s	Acceptable
Llama 3.2 1B Instruct	Q8_0	45 tok/s	Good
Llama 3.2 3B Instruct	Q8_0	35 tok/s	Good
Llama 3.3 70B Instruct	Q2_K	–	Not viable
LLaVA 1.6 13B	Q4_K_M	8 tok/s	Acceptable
Mistral 7B Instruct v0.3	Q4_K_M	14 tok/s	Acceptable
Mistral Small 24B Instruct	Q3_K_M	–	Not viable
Mixtral 8x7B Instruct	Q2_K	–	Not viable
nomic-embed-text v1.5	Q8_0	600 tok/s	Good
NVIDIA Nemotron-3-super-120B-A12B	Q2_K	–	Not viable
Phi-3 Medium 14B Instruct	Q3_K_M	6 tok/s	Marginal
Phi-3 Mini 3.8B Instruct	Q8_0	32 tok/s	Good
Phi-4 14B	Q3_K_M	5 tok/s	Marginal
Phi-4 Mini	Q8_0	30 tok/s	Good
Qwen 2.5 14B Instruct	Q3_K_M	5 tok/s	Marginal
Qwen 2.5 72B Instruct	Q2_K	–	Not viable
Qwen 2.5 7B Instruct	Q4_K_M	16 tok/s	Acceptable
Qwen 2.5 Coder 32B Instruct	Q3_K_M	–	Not viable
Qwen 2.5 Coder 7B Instruct	Q4_K_M	15 tok/s	Acceptable
Qwen3-14B Instruct	Q8_0	2 tok/s	Marginal
Qwen3-30B-A3B	Q4_K_M	3 tok/s	Marginal
Qwen3-32B Instruct	Q3_K_M	2 tok/s	Marginal
Qwen3-8B Instruct	Q8_0	8 tok/s	Marginal
Qwen3.5-122B-A10B	Q3_K_M	–	Not viable
Qwen3.5-27B	Q4_K_M	16 tok/s	Acceptable
Qwen3.5-397B (MoE)	Q2_K	–	Not viable
Qwen3.6-27B	Q4_K_M	16 tok/s	Acceptable
Qwen3.6-35B-A3B	Q3_K_M	3 tok/s	Marginal
QwQ 32B Preview	Q3_K_M	–	Not viable
Stable Diffusion 3 Medium	FP16	–	Good
Stable Diffusion 3.5 Large	FP16	–	Marginal
Stable Diffusion XL 1.0	FP16	–	Good
StarCoder 2 15B	Q3_K_M	4 tok/s	Marginal
Whisper Large V3	Q5_K_M	–	Good
Whisper Large V3 Turbo	FP16	–	Good
Yi 1.5 34B Chat	Q3_K_M	–	Not viable

Showing 59 of 59 entries

Buy Used Mac

Prices and availability vary. Inspect hardware before purchasing. Some links may be affiliate links.

eBay Swappa

FAQ

Frequently Asked Questions

What AI models can Apple M3 Pro (18GB Unified) run?: The Apple M3 Pro (18GB Unified) can run 59 AI models. Top performers include all-MiniLM-L6-v2, nomic-embed-text v1.5, Gemma 4 26B-A4B. See the full compatibility table above for speeds and quality ratings.
Is Apple M3 Pro (18GB Unified) good for AI coding?: Yes. With 18 GB, the Apple M3 Pro (18GB Unified) handles single-model coding workflows well at the Capable tier.
How much VRAM does Apple M3 Pro (18GB Unified) have?: The Apple M3 Pro (18GB Unified) has 18 GB of unified memory with 150 GB/s bandwidth.
Can Apple M3 Pro (18GB Unified) run 70B models?: 70B models can run on the Apple M3 Pro (18GB Unified) with CPU offloading, but performance will be reduced. Consider a GPU with 48GB+ VRAM for full-speed 70B inference.
Is Apple M3 Pro (18GB Unified) worth it for AI?: At $1,799, the Apple M3 Pro (18GB Unified) offers 18 GB VRAM and runs 59 AI models. It works for smaller models and experimentation.

Own this GPU?

See every AI model it supports, expected performance, and how to build around it.

Check my rig

All GPUs