FitMyLLM
APPLE· M2 Pro

Apple M2 Pro (16GB)

Running LLMs on the Apple M2 Pro (16GB) — the long read: which models fit at which quantisation, and the settings worth changing. · Or what else $1,299 buys

Usable memory
10.7 GB
Entry-level
Bandwidth
200
GB/s
Models Q4
248/498
50%
7B Q4 speed
~25
Good
Model coverage at Q450% of all
Estimated speed· By model size at Q4

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

3B
~59
tok/s
7B
~25
tok/s
14B
~13
tok/s
32B
—
18.0GB needed
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
10.7 GB of 16
Bandwidth
200 GB/s
FP16 Compute
6.8 TFLOPS
TDP
30W
Memory
Unified
Architecture
M2 Pro
MSRP
$1,299
123
Fast models · >30 tok/s
Real-time chat speed
248
Usable · >10 tok/s
Comfortable for all tasks
248
Total compatible
Fit in usable memory at Q4
Don’t want to buy?

Test M2 Pro (16GB) (or anything bigger) without committing. Pay by the second, cancel anytime.

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· 248
S
Falcon-H1R Tiny 90M0.09B
Falcon·256K ctx· chat· reasoning
1975
tok/s · 5% mem
›
S
SmolLM2 135M0.135B
Smollm·2K ctx· chat
1317
tok/s · 5% mem
›
S
nomic-embed-text-v1.5 100M0.14B
Embedding·8K ctx· chat
1270
tok/s · 5% mem
›
S
GPT-2 124M0.14B
Gpt2·1K ctx· chat
1270
tok/s · 5% mem
›
S
SmolVLM 256M0.256B
Smollm·8K ctx· chat· vision
694
tok/s · 6% mem
›
S
Gemma 3 270M0.27B
Gemma·32K ctx· chat
658
tok/s · 6% mem
›
S
Snowflake Arctic Embed M v2.00.305B
Embedding·8K ctx· embedding· multilingual
583
tok/s · 6% mem
›
S
bge-large-en-v1.5 335M0.335B
Embedding·1K ctx· chat
531
tok/s · 7% mem
›
S
mxbai-embed-large-v10.335B
Embedding·1K ctx· embedding
531
tok/s · 7% mem
›
S
Snowflake Arctic Embed L0.335B
Embedding·1K ctx· embedding
531
tok/s · 7% mem
›
S
LFM2 350M0.35B
Lfm·125K ctx· chat· tool_use
508
tok/s · 7% mem
›
S
SmolLM2 360M0.36B
Smollm·8K ctx· chat
494
tok/s · 7% mem
›
S
GPT-2 Medium 345M0.38B
Gpt2·1K ctx· chat
468
tok/s · 7% mem
›
S
Qwen 2.5 0.5B0.5B
Qwen·32K ctx· chat
356
tok/s · 7% mem
›
S
SmolVLM 500M0.5B
Smollm·8K ctx· chat· vision
356
tok/s · 7% mem
›
S
Falcon-H1 0.5B0.52B
Falcon·128K ctx· chat
342
tok/s · 8% mem
›
S
BGE-M30.568B
Embedding·8K ctx· embedding
313
tok/s · 8% mem
›
S
Snowflake Arctic Embed L v2.00.568B
Embedding·8K ctx· embedding· multilingual
313
tok/s · 8% mem
›
S
Qwen3-Embedding 0.6B0.6B
Embedding·32K ctx· embedding· multilingual
296
tok/s · 8% mem
›
S
Falcon-H1R Tiny 0.6B0.6B
Falcon·32K ctx· chat· reasoning
296
tok/s · 8% mem
›
S
Falcon Perception 0.6B0.6B
Falcon·4K ctx· vision
296
tok/s · 8% mem
›
S
Qwen 1.5 0.5B0.62B
Qwen·32K ctx· chat
287
tok/s · 8% mem
›
S
LFM2 700M0.74B
Lfm·125K ctx· chat· tool_use
240
tok/s · 9% mem
›
S
Qwen3 0.6B0.75B
Qwen·32K ctx· chat· reasoning
237
tok/s · 9% mem
›
S
GPT-2 Large 774M0.81B
Gpt2·1K ctx· chat
219
tok/s · 9% mem
›
S
Qwen 3.5 0.8B0.87B
Qwen·256K ctx· chat· coding· multilingual
204
tok/s · 10% mem
›
S
InternVL3 1B0.94B
Other·32K ctx· chat· vision
189
tok/s · 10% mem
›
S
MiniCPM5 1B1.08B
Minicpm·128K ctx· chat· reasoning· multilingual
165
tok/s · 11% mem
›
S
K2 Horizon 0.9B1.08B
Other·128K ctx· chat· coding· reasoning
165
tok/s · 11% mem
›
S
TinyLlama 1.1B1.1B
Llama·2K ctx· chat
162
tok/s · 11% mem
›
S
LFM2.5-1.2B-Thinking1.2B
Lfm·122K ctx· chat· reasoning· tool_use
148
tok/s · 11% mem
›
S
Llama-3.2-1B1.2B
Llama·4K ctx· chat
148
tok/s · 11% mem
›
S
LFM2 1.2B1.2B
Lfm·125K ctx· chat· tool_use· multilingual
148
tok/s · 11% mem
›
S
Zamba2 1.2B1.2B
Other·4K ctx· chat
148
tok/s · 11% mem
›
S
Ling 3.0 Tiny7.89BMoE
Other·128K ctx· chat· coding· reasoning
137
tok/s · 50% mem
›
S
EXAONE-4.0-1.2B1.3B
Exaone·64K ctx· chat
137
tok/s · 12% mem
›
S
OPT 1.3B1.3B
Opt·2K ctx· chat
137
tok/s · 12% mem
›
S
MiniCPM-V 4.61.3B
Other·256K ctx· chat· vision
137
tok/s · 12% mem
›
S
DeepSeek Coder 1.3B1.35B
Deepseek·16K ctx· coding
132
tok/s · 12% mem
›
S
Phi-1 1.3B1.42B
Phi·2K ctx· coding
125
tok/s · 13% mem
›
Next step

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Apple M2 Pro (16GB).

What this card is worth

M2 Pro (16GB) holds 248 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is Phi-3-medium-14b at IQ4_XS.

Tokens/sec per $100
2.2
8B at Q4_K_M, so cards compare like for like
Usable memory per $100
0.82 GB
what memory costs on this card
The biggest it takes
Phi-3-medium-14b14B · IQ4_XS9.4 GB17 tok/sest
Ministral 3 14B Reasoning13.95B · IQ4_XS9.3 GB17 tok/sest
Mistral-Nemo 12.2B12.2B · Q4_K_M9.0 GB17 tok/sest
Mellum2.1 12B-A2.5B Thinking12.15B · Q5_K_S9.6 GB75 tok/sest
Dolly v2 12B12B · Q4_K_M9.1 GB18 tok/sest
StableLM 2 12B12B · Q4_K_M9.0 GB18 tok/sest
Falcon2 11B11B · Q5_K_S9.5 GB17 tok/sest
Llama-3.2-11B-Vision-Instruct11B · Q4_K_S9.4 GB20 tok/sest
Against cards of similar memory
M3 Pro (18GB)12 GB$1,5991.5 tok/s per $100
M2 Pro (16GB)11 GB$1,2992.2 tok/s per $100
M1 Pro (16GB)11 GB$9992.9 tok/s per $100
M4 (16GB)11 GB$4994.0 tok/s per $100
M2 (16GB)11 GB$7992.1 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 M2 Pro (16GB) — 16 GB unified, 10.7 GB usable.

Apple M2 Pro (16GB) spec
Brand
Apple
Unified memory
16 GB
Usable by a model
10.7 GB
Bandwidth
200 GB/s
FP16 Compute
6.8 TFLOPS
FP32 Compute
6.8 TFLOPS
TDP
30 W
Architecture
M2 Pro
MSRP
$1299
AI capability
248/ 498 models @ Q4

With 10.7 GB of its 16 GB reaching a model and 200 GB/s bandwidth, this machine handles models up to 13B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~23 tok/s.

Rent it — live prices
Checking what the clouds are charging…
Top models for Apple M2 Pro (16GB)
248 fit · showing 20
ModelSizeVRAM Q4tok/sAvg
Baichuan2 13B13B8.4 GB1423.6
Llama 2 13B13B8.4 GB1417.2
CodeLlama 13B13B8.4 GB1419.7
Vicuna 13B13B8.4 GB1411.8
LLaMA 1 13B13B8.4 GB1432.9
OPT 13B13B8.4 GB1435.8
Orca 2 13B13B8.4 GB1425.4
WizardCoder Python 13B13B8.4 GB1460.1
WizardLM 13B13B8.4 GB1419.5
Mistral-Nemo 12.2B12.2B7.9 GB15129.3
Mellum2.1 12B-A2.5B Thinking12.15B7.9 GB7164.5
Dolly v2 12B12B7.8 GB156.4
StableLM 2 12B12B7.8 GB1521.3
Falcon2 11B11B7.2 GB1633.2
SOLAR-10.7B10.7B7.0 GB1728.2
Falcon3-10B10.3B6.8 GB1738.2
GLM-4.1V 9B Thinking10.29B6.8 GB1770.3
Bamba 9B v29.78B6.5 GB1826.1
Qwen 3.5 9B9.65B6.4 GB1849.2
RecurrentGemma 9B9.63B6.4 GB1835.0