FitMyLLM
APPLE· M2 ULTRA

Apple M2 Ultra (192GB)

Running LLMs on the Apple M2 Ultra (192GB) — the long read: which models fit at which quantisation, and the settings worth changing. · Or what else $5,499 buys

USABLE MEMORY
144 GB
FLAGSHIP
BANDWIDTH
800
GB/S
MODELS Q4
398/449
89%
7B Q4 SPEED
~102
BLAZING
▸ MODEL COVERAGE @ Q489% OF ALL
▸ ESTIMATED SPEED· BY MODEL SIZE @ Q4

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

3B
~237
TOK/S
7B
~102
TOK/S
14B
~51
TOK/S
32B
~22
TOK/S
70B
~10
TOK/S
▸ 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
144 GB of 192
BANDWIDTH
800 GB/s
FP16 COMPUTE
27.2 TFLOPS
TDP
215W
MEMORY
LPDDR5
ARCHITECTURE
M2 Ultra
MSRP
$5,499
310
FAST MODELS · >30 TOK/S
Real-time chat speed
384
USABLE · >10 TOK/S
Comfortable for all tasks
398
TOTAL COMPATIBLE
Fit in usable memory at Q4
▸ DON’T WANT TO BUY?

Test M2 Ultra (192GB) (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· 398
S
Falcon-H1R Tiny 90M0.09B
FALCON·256K CTX· CHAT· REASONING
7901
TOK/S · 0% MEM
S
SmolLM2 135M0.135B
SMOLLM·2K CTX· CHAT
5267
TOK/S · 0% MEM
S
nomic-embed-text-v1.5 100M0.14B
EMBEDDING·8K CTX· CHAT
5079
TOK/S · 0% MEM
S
GPT-2 124M0.14B
GPT2·1K CTX· CHAT
5079
TOK/S · 0% MEM
S
SmolVLM 256M0.256B
SMOLLM·8K CTX· CHAT· VISION
2778
TOK/S · 0% MEM
S
Gemma 3 270M0.27B
GEMMA·32K CTX· CHAT
2634
TOK/S · 0% MEM
S
Snowflake Arctic Embed M v2.00.305B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
2332
TOK/S · 0% MEM
S
bge-large-en-v1.5 335M0.335B
EMBEDDING·1K CTX· CHAT
2123
TOK/S · 0% MEM
S
mxbai-embed-large-v10.335B
EMBEDDING·1K CTX· EMBEDDING
2123
TOK/S · 0% MEM
S
Snowflake Arctic Embed L0.335B
EMBEDDING·1K CTX· EMBEDDING
2123
TOK/S · 0% MEM
S
LFM2 350M0.35B
LFM·125K CTX· CHAT· TOOL_USE
2032
TOK/S · 0% MEM
S
SmolLM2 360M0.36B
SMOLLM·8K CTX· CHAT
1975
TOK/S · 0% MEM
S
GPT-2 Medium 345M0.38B
GPT2·1K CTX· CHAT
1871
TOK/S · 1% MEM
S
Qwen 2.5 0.5B0.5B
QWEN·32K CTX· CHAT
1422
TOK/S · 1% MEM
S
SmolVLM 500M0.5B
SMOLLM·8K CTX· CHAT· VISION
1422
TOK/S · 1% MEM
S
Falcon-H1 0.5B0.52B
FALCON·128K CTX· CHAT
1368
TOK/S · 1% MEM
S
BGE-M30.568B
EMBEDDING·8K CTX· EMBEDDING
1252
TOK/S · 1% MEM
S
Snowflake Arctic Embed L v2.00.568B
EMBEDDING·8K CTX· EMBEDDING· MULTILINGUAL
1252
TOK/S · 1% MEM
S
Qwen3-Embedding 0.6B0.6B
EMBEDDING·32K CTX· EMBEDDING· MULTILINGUAL
1185
TOK/S · 1% MEM
S
Falcon-H1R Tiny 0.6B0.6B
FALCON·32K CTX· CHAT· REASONING
1185
TOK/S · 1% MEM
S
Falcon Perception 0.6B0.6B
FALCON·4K CTX· VISION
1185
TOK/S · 1% MEM
S
Qwen 1.5 0.5B0.62B
QWEN·32K CTX· CHAT
1147
TOK/S · 1% MEM
S
LFM2 700M0.74B
LFM·125K CTX· CHAT· TOOL_USE
961
TOK/S · 1% MEM
S
Qwen3 0.6B0.75B
QWEN·32K CTX· CHAT· REASONING
948
TOK/S · 1% MEM
S
GPT-2 Large 774M0.81B
GPT2·1K CTX· CHAT
878
TOK/S · 1% MEM
S
Qwen 3.5 0.8B0.87B
QWEN·256K CTX· CHAT· CODING· MULTILINGUAL
817
TOK/S · 1% MEM
S
InternVL3 1B0.94B
OTHER·32K CTX· CHAT· VISION
757
TOK/S · 1% MEM
S
MiniCPM5 1B1.08B
MINICPM·128K CTX· CHAT· REASONING· MULTILINGUAL
658
TOK/S · 1% MEM
S
TinyLlama 1.1B1.1B
LLAMA·2K CTX· CHAT
646
TOK/S · 1% MEM
S
LFM2.5-1.2B-Thinking1.2B
LFM·122K CTX· CHAT· REASONING· TOOL_USE
593
TOK/S · 1% MEM
S
Llama-3.2-1B1.2B
LLAMA·4K CTX· CHAT
593
TOK/S · 1% MEM
S
LFM2 1.2B1.2B
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
593
TOK/S · 1% MEM
S
Zamba2 1.2B1.2B
OTHER·4K CTX· CHAT
593
TOK/S · 1% MEM
S
EXAONE-4.0-1.2B1.3B
EXAONE·64K CTX· CHAT
547
TOK/S · 1% MEM
S
OPT 1.3B1.3B
OPT·2K CTX· CHAT
547
TOK/S · 1% MEM
S
MiniCPM-V 4.61.3B
OTHER·256K CTX· CHAT· VISION
547
TOK/S · 1% MEM
S
DeepSeek Coder 1.3B1.35B
DEEPSEEK·16K CTX· CODING
527
TOK/S · 1% MEM
S
Phi-1 1.3B1.42B
PHI·2K CTX· CODING
501
TOK/S · 1% MEM
S
Phi-1.5 1.3B1.42B
PHI·2K CTX· CHAT· CODING
501
TOK/S · 1% MEM
S
LFM2 8B A1B8.3BMoE
LFM·125K CTX· CHAT· TOOL_USE· MULTILINGUAL
474
TOK/S · 4% MEM
▸ NEXT STEP

Get personalized recommendations.

See ranked models with benchmark scores, run commands, and precise speed estimates for your Apple M2 Ultra (192GB).

WHAT THIS CARD IS WORTH

M2 Ultra (192GB) holds 398 of the models in our catalogue and is, in practice, a IQ4_XS card — the largest it takes is Command A+ at IQ4_XS.

TOKENS/SEC PER $100
1.3
8B at Q4_K_M, so cards compare like for like
USABLE MEMORY PER $100
2.62 GB
what memory costs on this card
THE BIGGEST IT TAKES
Command A+218B · IQ4_XS124.5 GB23 tok/sEST
Step 3.7 Flash201.37B · Q4_K_M125.8 GB48 tok/sEST
Falcon 180B180B · Q5_K_S129.1 GB3 tok/sEST
bloom 176.2B176.2B · Q4_K_M125.5 GB3 tok/sEST
dots.llm1.inst 142.8B142.8B · Q6_K123.5 GB3 tok/sEST
WizardLM 2 8x22B141B · Q6_K119.0 GB10 tok/sEST
Mixtral-8x22B140.6B · Q6_K118.7 GB10 tok/sEST
DBRX 132B132B · Q6_K111.4 GB11 tok/sEST
AGAINST CARDS OF SIMILAR MEMORY
Radeon Instinct MI300A192 GB$12,0009.2 tok/s per $100
Radeon Instinct MI308X192 GB$12,0009.2 tok/s per $100
Instinct MI300X192 GB$15,0003.8 tok/s per $100
M4 Ultra (192GB)144 GB$7,4991.2 tok/s per $100
M2 Ultra (192GB)144 GB$5,4991.3 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 Ultra (192GB) — 192 GB unified, 144 GB usable.

APPLE M2 ULTRA (192GB) SPEC
BRAND
Apple
UNIFIED MEMORY
192 GB
USABLE BY A MODEL
144 GB
BANDWIDTH
800 GB/s
FP16 COMPUTE
27.2 TFLOPS
FP32 COMPUTE
27.2 TFLOPS
TDP
215 W
ARCHITECTURE
M2 Ultra
MSRP
$5499
▸ AI CAPABILITY
398/ 449 models @ Q4

With 144 GB of its 192 GB reaching a model and 800 GB/s bandwidth, this machine handles models up to 201.37B parameters.

Speed ≈ bandwidth / model_size × efficiency. A 7B model at Q4 runs at ~91 tok/s.

RENT IT — LIVE PRICES
Checking what the clouds are charging…
§ 01TOP MODELS FOR APPLE M2 ULTRA (192GB)
398 FIT · SHOWING 20
MODELSIZEVRAM Q4TOK/SAVG
Step 3.7 Flash201.37B123.6 GB6553.8
Falcon 180B180B110.5 GB451.3
bloom 176.2B176.2B108.2 GB415.0
dots.llm1.inst 142.8B142.8B87.8 GB5
WizardLM 2 8x22B141B86.7 GB1842.4
Mixtral-8x22B140.6B86.4 GB1831.9
DBRX 132B132B81.2 GB2046.3
Mistral Medium 3.5128B78.7 GB646.6
Pixtral Large 124B124B76.3 GB639.3
Nemotron 3 Super 120B-A12B123.61B76.0 GB5953.2
Mistral-Large 123B123B75.7 GB633.5
Devstral 2 123B123B75.7 GB638.1
Qwen 3.5 122B A10B122B75.1 GB7156.8
Nemotron 3 Super 120B120B73.8 GB5957.3
Mistral Small 4 119B119B73.2 GB10950.2
GPT-OSS 120B117B72.0 GB13954.1
Command A 111B111B68.3 GB627.6
GLM 4.5 Air110B67.7 GB5951.0
Qwen 1.5 110B110B67.7 GB633.4
Llama 4 Scout 17B-16E109B67.1 GB4233.9