FitMyLLM
APPLE· M3

Apple M3 (8GB)

Running LLMs on the Apple M3 (8GB) — the long read: which models fit at which quantisation, and the settings worth changing. · Or what else $599 buys

USABLE MEMORY
5 GB
BUDGET
BANDWIDTH
100
GB/S
MODELS Q4
104/449
23%
7B Q4 SPEED
~13
USABLE
▸ MODEL COVERAGE @ Q423% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

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

3B
~30
TOK/S
7B
~13
TOK/S
14B
7.9GB NEEDED
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
5 GB of 8
BANDWIDTH
100 GB/s
FP16 COMPUTE
4.1 TFLOPS
TDP
22W
MEMORY
Unified
ARCHITECTURE
M3
MSRP
$599
80
FAST MODELS · >30 TOK/S
Real-time chat speed
104
USABLE · >10 TOK/S
Comfortable for all tasks
104
TOTAL COMPATIBLE
Fit in usable memory at Q4
▸ DON’T WANT TO BUY?

Test M3 (8GB) (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· 104
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
988
TOK/S · 11% MEM
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
658
TOK/S · 11% MEM
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
635
TOK/S · 11% MEM
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
635
TOK/S · 11% MEM
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
347
TOK/S · 13% MEM
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
329
TOK/S · 13% MEM
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
291
TOK/S · 13% MEM
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
265
TOK/S · 14% MEM
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
265
TOK/S · 14% MEM
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
265
TOK/S · 14% MEM
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
254
TOK/S · 14% MEM
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
247
TOK/S · 14% MEM
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
234
TOK/S · 14% MEM
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
178
TOK/S · 16% MEM
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
178
TOK/S · 16% MEM
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
171
TOK/S · 16% MEM
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
156
TOK/S · 17% MEM
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
156
TOK/S · 17% MEM
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
148
TOK/S · 17% MEM
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
148
TOK/S · 17% MEM
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
148
TOK/S · 17% MEM
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
143
TOK/S · 17% MEM
S
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
120
TOK/S · 19% MEM
S
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
119
TOK/S · 19% MEM
S
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
110
TOK/S · 20% MEM
S
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
102
TOK/S · 20% MEM
S
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
95
TOK/S · 21% MEM
S
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
82
TOK/S · 23% MEM
S
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
81
TOK/S · 23% MEM
S
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
74
TOK/S · 24% MEM
S
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
74
TOK/S · 24% MEM
S
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
74
TOK/S · 24% MEM
S
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
74
TOK/S · 24% MEM
S
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
68
TOK/S · 26% MEM
S
OPT 1.3B1.3B
OPT·2K CTX· CHAT
68
TOK/S · 26% MEM
S
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
68
TOK/S · 26% MEM
S
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
66
TOK/S · 26% MEM
S
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
63
TOK/S · 27% MEM
S
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
63
TOK/S · 27% MEM
A
Qwen2.5-Coder-1.5B1.5B
QWEN·32K CTX· CHAT· TOOL_USE· CODING
59
TOK/S · 28% MEM
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Apple M3 (8GB).

WHAT THIS CARD IS WORTH

M3 (8GB) holds 104 of the models in our catalogue and is, in practice, a Q4_K_M card — the largest it takes is ChatGLM2 6B at Q4_K_M.

TOKENS/SEC PER $100
2.8
8B at Q4_K_M, so cards compare like for like
USABLE MEMORY PER $100
0.83 GB
what memory costs on this card
THE BIGGEST IT TAKES
ChatGLM2 6B6.24B · Q4_K_M4.4 GB20 tok/sEST
ChatGLM3 6B6.24B · Q4_K_M4.4 GB20 tok/sEST
Yi-1.5 6B6.06B · Q4_K_M4.4 GB20 tok/sEST
Yi 6B6B · Q4_K_M4.4 GB20 tok/sEST
Gemma 4 E2B5.1B · Q4_K_M4.4 GB24 tok/sEST
Qwen 3.5 4B4.66B · Q4_K_M4.1 GB26 tok/sEST
InternLM2 5B4.5B · Q5_K_M4.1 GB23 tok/sEST
Qwen3-VL 4B Instruct4.44B · Q4_K_M4.4 GB28 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
RTX 3070 Ti8 GB$49912.6 tok/s per $100
RTX 3060 Ti GDDR6X8 GB$39915.8 tok/s per $100
RTX 3070 Ti 8 GB GA1028 GB$59910.5 tok/s per $100
Arc A7508 GB$19910.1 tok/s per $100
M3 (8GB)5 GB$5992.8 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 (8GB) — 8 GB unified, 5 GB usable.

APPLE M3 (8GB) SPEC
BRAND
Apple
UNIFIED MEMORY
8 GB
USABLE BY A MODEL
5 GB
BANDWIDTH
100 GB/s
FP16 COMPUTE
4.1 TFLOPS
FP32 COMPUTE
4.1 TFLOPS
TDP
22 W
ARCHITECTURE
M3
MSRP
$599
▸ AI CAPABILITY
104/ 449 models @ Q4

With 5 GB of its 8 GB reaching a model and 100 GB/s bandwidth, this machine handles models up to 4.5B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~11 tok/s.

RENT IT — LIVE PRICES
Checking what the clouds are charging…
§ 01TOP MODELS FOR APPLE M3 (8GB)
104 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
InternLM2 5B4.5B3.2 GB2047.6
Qwen3-VL 4B Instruct4.44B3.2 GB2026.6
MedGemma 1.5 4B4.3B3.1 GB215.5
Gemma 3 4B4.3B3.1 GB2122.8
Ministral 3 3B Reasoning4.25B3.1 GB21
Qwen 1.5 4B4B2.9 GB2212.6
Qwen3 4B4B2.9 GB2240.7
Qwen3-4B Instruct 25074B2.9 GB2237.2
Qwen3-Embedding 4B4B2.9 GB22
Granite Vision 4.1 4B4B2.9 GB22
Nemotron 3 Nano 4B3.97B2.9 GB2232.0
Ministral 3 3B3.85B2.8 GB2321.4
Phi-3.5 Mini 3.8B3.82B2.8 GB2346.6
phi-3-mini-4k 3.8B3.8B2.8 GB2330.5
Phi-4-mini 3.8B3.8B2.8 GB2349.0
Qwen2.5-VL-3B3.8B2.8 GB2329.9
Jina Embeddings v43.8B2.8 GB23
Cogito 3B3.61B2.7 GB2522.1
Falcon3-3B3.23B2.5 GB2825.7
granite-4.0-h-micro 3.2B3.2B2.4 GB2818.4