▸ APPLE· M3 PRO

Apple M3 Pro (36GB)

Running LLMs on the Apple M3 Pro (36GB) — the long read: which models fit at which quantisation, and the settings worth changing. · Or what else $1,999 buys

USABLE MEMORY

27 GB

HIGH-END

BANDWIDTH

150

GB/S

MODELS Q4

330/449

73%

7B Q4 SPEED

~19

GOOD

▸ MODEL COVERAGE @ Q473% OF ALL

▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

Average speeds at Q4 quantization. Actual performance varies by model architecture and context length.

~44

TOK/S

~19

TOK/S

14B

~10

TOK/S

32B

TOK/S

70B

—

39.4GB NEEDED

▸ MEASURED RIG REPORTS

We rent the machine and time every model on it: decode, VRAM peak, concurrency, watts and cost per million tokens. Yours may already be one of them — and two are free to read in full.

SEE THE REPORTS →

▸ SPECIFICATIONS

USABLE MEMORY

27 GB of 36

BANDWIDTH

150 GB/s

FP16 COMPUTE

14.2 TFLOPS

TDP

30W

MEMORY

LPDDR5

ARCHITECTURE

M3 Pro

MSRP

$1,999

124

FAST MODELS · >30 TOK/S

Real-time chat speed

263

USABLE · >10 TOK/S

Comfortable for all tasks

330

TOTAL COMPATIBLE

Fit in usable memory at Q4

▸ DON’T WANT TO BUY?

Test M3 Pro (36GB) (or anything bigger) without committing. Pay by the second, cancel anytime.

Rent on Vast.ai →Or RunPod →

Spin up in ~60s. Pay by the second. Cancel anytime.

Some links are affiliate links — we may earn a small commission at no extra cost to you. This helps keep FitMyLLM free and independent.

▸ COMPATIBLE MODELS· 330

Falcon-H1R Tiny 90M0.09B

FALCON·256K CTX· CHAT· REASONING

nomic-embed-text-v1.5 100M0.14B

EMBEDDING·8K CTX· CHAT

SMOLLM·8K CTX· CHAT· VISION

Snowflake Arctic Embed M v2.00.305B

EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL

bge-large-en-v1.5 335M0.335B

EMBEDDING·1K CTX· CHAT

mxbai-embed-large-v10.335B

EMBEDDING·1K CTX· EMBEDDING

Snowflake Arctic Embed L0.335B

EMBEDDING·1K CTX· EMBEDDING

LFM·125K CTX· CHAT· TOOL_USE

GPT-2 Medium 345M0.38B

SMOLLM·8K CTX· CHAT· VISION

FALCON·128K CTX· CHAT

EMBEDDING·8K CTX· EMBEDDING

Snowflake Arctic Embed L v2.00.568B

EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL

Qwen3-Embedding 0.6B0.6B

EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL

Falcon-H1R Tiny 0.6B0.6B

FALCON·32K CTX· CHAT· REASONING

Falcon Perception 0.6B0.6B

FALCON·4K CTX· VISION

LFM·125K CTX· CHAT· TOOL_USE

QWEN·32K CTX· CHAT· REASONING

GPT-2 Large 774M0.81B

QWEN·256K CTX· CHAT· CODING· MULTILINGUAL

OTHER·32K CTX· CHAT· VISION

MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL

LFM2.5-1.2B-Thinking1.2B

LFM·122K CTX· CHAT· REASONING· TOOL_USE

LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL

OTHER·256K CTX· CHAT· VISION

DeepSeek Coder 1.3B1.35B

DEEPSEEK·16K CTX· CODING

PHI·2K CTX· CHAT· CODING

LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL

▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Apple M3 Pro (36GB).

▸Find models for Apple M3 Pro (36GB)Compare with other GPUs

WHAT THIS CARD IS WORTH

M3 Pro (36GB) holds 330 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is Seed-OSS 36B Instruct at IQ4_XS.

TOKENS/SEC PER $100

1.2

8B at Q4_K_M, so cards compare like for like

USABLE MEMORY PER $100

1.35 GB

what memory costs on this card

THE BIGGEST IT TAKES

Seed-OSS 36B Instruct36B · IQ4_XS23.5 GB5 tok/sEST

c4ai-command-r-v01 35B35B · Q4_K_S23.5 GB5 tok/sEST

Qwen 3.5 35B A3B35B · Q4_K_M24.0 GB59 tok/sEST

Qwen 3.6 35B A3B35B · Q4_K_M24.0 GB59 tok/sEST

Nous Capybara 34B34.4B · Q4_K_S23.4 GB5 tok/sEST

Yi-1.5 34B34.4B · Q4_K_S23.4 GB5 tok/sEST

Falcon-H1 34B34B · Q4_K_M23.8 GB5 tok/sEST

CodeLlama 34B34B · Q4_K_M24.0 GB5 tok/sEST

AGAINST CARDS OF SIMILAR MEMORY

M4 Max (48GB)36 GB$2,4992.2 tok/s per $100

M3 Max (48GB)36 GB$2,8991.6 tok/s per $100

M4 Pro (48GB)36 GB$1,7991.9 tok/s per $100

M3 Max (36GB)27 GB$2,4991.5 tok/s per $100

M3 Pro (36GB)27 GB$1,9991.2 tok/s per $100

Neighbours in memory rather than in price: memory decides whether a card can do the job at all, so two cards of the same size at different prices is the comparison you are making. Ordered by memory, then bandwidth — not by the value column, which is worked out from bandwidth and price alone and therefore rewards a cheap card whatever its software stack does to that bandwidth in practice. Read it as one input, not as a ranking.

▸ DEVICE UNDER TEST

Apple M3 Pro (36GB) — 36 GB unified, 27 GB usable.

▸ APPLE M3 PRO (36GB) SPEC

BRAND: Apple
UNIFIED MEMORY: 36 GB
USABLE BY A MODEL: 27 GB
BANDWIDTH: 150 GB/s
FP16 COMPUTE: 14.2 TFLOPS
FP32 COMPUTE: 14.2 TFLOPS
TDP: 30 W
ARCHITECTURE: M3 Pro
MSRP: $1999

▸ AI CAPABILITY

330/ 449 models @ Q4

With 27 GB of its 36 GB reaching a model and 150 GB/s bandwidth, this machine handles models up to 34.4B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~17 tok/s.

RENT IT — LIVE PRICES

Checking what the clouds are charging…

§ 01TOP MODELS FOR APPLE M3 PRO (36GB)

330 FIT · SHOWING 20

MODEL	SIZE	VRAM Q4	TOK/S	AVG
Nous Capybara 34B	34.4B	21.5 GB	4	42.0
Yi-1.5 34B	34.4B	21.5 GB	4	45.3
Falcon-H1 34B	34B	21.3 GB	4	66.1
CodeLlama 34B	34B	21.3 GB	4	25.4
Nous Hermes 2 34B	34B	21.3 GB	4	47.0
Phind CodeLlama 34B	34B	21.3 GB	4	68.1
LLaVA-1.6 Yi 34B	34B	21.3 GB	4	47.4
WizardCoder Python 34B	34B	21.3 GB	4	73.2
Yi 34B	34B	21.3 GB	4	33.4
Qwen3-VL 32B Instruct	33.36B	20.9 GB	4	44.6
DeepSeek Coder 33B	33B	20.7 GB	4	26.0
Vicuna 33B	33B	20.7 GB	4	17.2
LLaMA 1 30B	33B	20.7 GB	4	17.8
DeepSeek-R1-Distill-Qwen-32B	32.8B	20.5 GB	4	46.9
Qwen3 32B	32.8B	20.5 GB	4	54.9
Gemma 4 31B	32.68B	20.5 GB	4	53.7
Qwen2.5-32B	32.5B	20.4 GB	4	54.3
Qwen 2.5 Coder 32B	32.5B	20.4 GB	4	48.0
QwQ-32B	32.5B	20.4 GB	4	43.7
Granite Switch 4.1 30B Preview	32.24B	20.2 GB	4	36.1