Apple · M2 · unified memory

What runs on a Mac Studio M2 Ultra 192GB

192 GB at 800 GB/s. The largest model that fits at Q4_K_M is Qwen3 235B-A22B, at roughly 23.9 tokens per second. Unified memory means the operating system takes a share, so budget around three quarters of the total.

Runs comfortably

6.1 GB of 144.0 GB · 4%
0144 GB
Weights 4.6 GB
KV cache 1.0 GB
Runtime overhead 0.5 GB

Llama 3.1 8B at Q4_K_M leaves 137.9 GB spare on a Mac Studio M2 Ultra 192GB. There is room to raise the context length or move up a quantisation level.

Generation87.5tok/s
Prompt processing1412tok/s
Max context128Ktokens
KV per 1K tokens0GB

Every quantisation of Llama 3.1 8B on a Mac Studio M2 Ultra 192GB

Highlighted row is the highest quality that still fits at 8K context.

QuantisationbpwWeights TotalVerdictMax ctxtok/s
FP16 / BF16 16.00 15.0 GB 16.5 GB Runs comfortably 128K 29.3
INT8 / W8A8 8.50 7.7 GB 9.2 GB Runs comfortably 128K 54.9
Q8_0 (GGUF) 8.50 7.7 GB 9.2 GB Runs comfortably 128K 54.9
FP8 (E4M3) 8.00 7.5 GB 9.0 GB Runs comfortably 128K 56.4
Q6_K 6.56 6.1 GB 7.7 GB Runs comfortably 128K 67.7
AWQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 75.9
GPTQ 4-bit 4.25 5.4 GB 6.9 GB Runs comfortably 128K 75.9
MXFP4 4.25 5.4 GB 6.9 GB Runs comfortably 128K 75.9
Q5_K_M 5.67 5.3 GB 6.8 GB Runs comfortably 128K 77.3
Q5_K_S 5.52 5.2 GB 6.7 GB Runs comfortably 128K 79.2
Q4_K_M 4.85 4.6 GB 6.1 GB Runs comfortably 128K 87.5
Q4_K_S 4.58 4.4 GB 5.9 GB Runs comfortably 128K 91.4
Q4_0 4.55 4.4 GB 5.9 GB Runs comfortably 128K 91.8
IQ4_XS 4.25 4.1 GB 5.6 GB Runs comfortably 128K 96.6
Q3_K_M 3.91 3.8 GB 5.4 GB Runs comfortably 128K 103
IQ3_M 3.70 3.7 GB 5.2 GB Runs comfortably 128K 107
IQ3_XXS 3.06 3.2 GB 4.7 GB Runs comfortably 128K 122
Q2_K 2.63 2.8 GB 4.3 GB Runs comfortably 128K 134
IQ2_XXS 2.06 2.3 GB 3.9 GB Runs comfortably 128K 155
IQ1_M 1.75 2.1 GB 3.6 GB Runs comfortably 128K 170

Models on a Mac Studio M2 Ultra 192GB

Q4_K_M at 8K context. Bandwidth sets the speed; capacity sets the ceiling.

ModelParamsWeightsVerdictMax ctxtok/s
DeepSeek-R1 671B-A37B 671B 379.0 GB Won't fit — 2.52
Qwen3 235B-A22B 235.1B 132.8 GB Fits, but tight 58K 23.9
gpt-oss 120B-A5.1B 116.8B 66.0 GB Runs comfortably 128K 104
Llama 4 Scout 109B-A17B 109B 61.7 GB Runs comfortably 436K 37.0
Qwen2.5 72B 72.7B 41.2 GB Runs comfortably 128K 10.7
Llama 3.1 70B 70.5B 40.0 GB Runs comfortably 128K 11.0
Mixtral 8x7B 46.7B 26.4 GB Runs comfortably 32K 48.6
Command R 35B 35.0B 20.1 GB Runs comfortably 99K 18.1
Yi-1.5 34B 34.4B 19.5 GB Runs comfortably 32K 22.1
Qwen3 32B 32.8B 18.6 GB Runs comfortably 128K 23.0
Qwen2.5-Coder 32B 32.8B 18.6 GB Runs comfortably 128K 23.0
DeepSeek-R1-Distill-Qwen 32B 32.8B 18.6 GB Runs comfortably 128K 23.0
Qwen3 30B-A3B 30.5B 17.3 GB Runs comfortably 128K 101
Gemma 3 27B 27.4B 15.6 GB Runs comfortably 128K 27.6
Mistral Small 24B 23.6B 13.4 GB Runs comfortably 32K 32.2
gpt-oss 20B-A3.6B 20.9B 11.9 GB Runs comfortably 128K 149
Qwen3 14B 14.8B 8.5 GB Runs comfortably 128K 49.5
Phi-4 14B 14.7B 8.4 GB Runs comfortably 16K 49.1
Mistral NeMo 12B 12.3B 7.0 GB Runs comfortably 128K 58.7
Gemma 3 12B 12.2B 7.0 GB Runs comfortably 128K 59.4
GLM-4 9B 9.4B 5.4 GB Runs comfortably 128K 80.3
Qwen3 8B 8.2B 4.7 GB Runs comfortably 128K 84.6
Llama 3.1 8B 8.0B 4.6 GB Runs comfortably 128K 87.5
DeepSeek-R1-Distill-Llama 8B 8.0B 4.6 GB Runs comfortably 128K 87.5
Qwen2.5 7B 7.6B 4.4 GB Runs comfortably 128K 97.3
Mistral 7B v0.3 7.3B 4.1 GB Runs comfortably 32K 96.8
Gemma 3 4B 4.3B 2.5 GB Runs comfortably 128K 164
Qwen3 4B 4.0B 2.3 GB Runs comfortably 128K 154
Llama 3.2 3B 3.2B 1.8 GB Runs comfortably 128K 193
Llama 3.2 1B 1.2B 0.7 GB Runs comfortably 128K 512

Direct answers

Llama 3.1 8B on Mac Studio M2 Ultra 192GB
8.0BLlama 3.1 70B on Mac Studio M2 Ultra 192GB
70.5BLlama 3.2 3B on Mac Studio M2 Ultra 192GB
3.2BLlama 3.2 1B on Mac Studio M2 Ultra 192GB
1.2BLlama 4 Scout 109B-A17B on Mac Studio M2 Ultra 192GB
109BQwen3 8B on Mac Studio M2 Ultra 192GB
8.2BQwen3 14B on Mac Studio M2 Ultra 192GB
14.8BQwen3 32B on Mac Studio M2 Ultra 192GB
32.8B