AI Compatibility Tier List
MacBook Pro M2 Max 96GB
374 AI models ranked by how well they run on your 96 GB unified memory hardware. Best match: Qwen 3.6 35B A3B (score: 92).
374
Models tested
329
Can run
96 GB
unified memory
32
Best tok/s
Tier Distribution
S 20
A 56
B 60
C 193
F 45
S: 20A: 56B: 60C: 193F: 45
S
S Tier -- Perfect Fit (20 models)
These models run flawlessly with headroom to spare.
Qwen 3.6 35B A3B35B36.7 GB32 tok/s142K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B31.3 GB35 tok/s256K ctx
moe
Qwen 3.5 35B A3B35B34.1 GB35 tok/s131K ctx
moe
Qwen3-VL 30B A3B Instruct30B31.0 GB36 tok/s256K ctx
moe
Qwen 3 30B A3B30.5B31.3 GB35 tok/s131K ctx
moe
Qwen3-Coder-Next80B61.5 GB17 tok/s99K ctx
moe
Qwen 3.5 27B27B30.9 GB15 tok/s131K ctx
dense
Qwen 3 32B32B34.7 GB13 tok/s131K ctx
dense
Qwen 3.5 9B9B19.0 GB45 tok/s131K ctx
dense
Magistral Small 250724B28.3 GB17 tok/s131K ctx
dense
Nemotron Cascade 2 30B A3B30B32.5 GB36 tok/s216K ctx
moe
Qwen 3.6 27B27B28.7 GB12 tok/s262K ctx
dense
Devstral Small 2 24B Instruct24B28.3 GB17 tok/s256K ctx
dense
GPT-OSS 20B21B26.5 GB45 tok/s128K ctx
moe
Nemotron 3 Nano 30B30B32.0 GB14 tok/s131K ctx
dense
Qwen 3 14B14B22.2 GB29 tok/s131K ctx
dense
Qwen 2.5 VL 72B72B60.1 GB6 tok/s33K ctx
dense
Qwen 3 8B8B18.3 GB51 tok/s131K ctx
dense
Gemma 4 31B30.7B44.6 GB10 tok/s43K ctx
dense
Qwen 3.5 4B4B15.9 GB56 tok/s131K ctx
dense
A
A Tier -- Great Fit (56 models)
Excellent performance with comfortable memory margins.
Phi-4-reasoning-plus 14B14.7B23.3 GB28 tok/s33K ctx
dense
Devstral Small 1.124B28.3 GB17 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B30.3 GB38 tok/s186K ctx
moe
Phi-4 Mini Reasoning 4B3.8B15.1 GB53 tok/s131K ctx
dense
EXAONE 4.0 32B32B34.7 GB13 tok/s131K ctx
dense
Nemotron Nano 8B8B18.1 GB51 tok/s131K ctx
dense
Ministral 3 14B14B22.2 GB29 tok/s262K ctx
multimodal
Qwen 2.5 32B32B34.7 GB13 tok/s131K ctx
dense
Gemma 3 27B27B39.0 GB11 tok/s59K ctx
dense
Llama 3.3 70B70B58.9 GB6 tok/s50K ctx
dense
Kimi Linear 48B A3B48B41.5 GB8 tok/s493K ctx
linear
Codestral 2 25.0822B27.1 GB18 tok/s256K ctx
dense
Mistral Small 3.2 24B24B28.3 GB17 tok/s131K ctx
vision
OLMo 2 32B32B34.7 GB13 tok/s4K ctx
dense
Granite 4.1 30B30B33.5 GB14 tok/s131K ctx
dense
LFM2 24B24B28.3 GB17 tok/s131K ctx
dense
CogVLM2 19B19B25.3 GB22 tok/s8K ctx
dense
InternVL2 8B8B18.1 GB51 tok/s8K ctx
dense
Mistral Small 24B24B28.3 GB17 tok/s33K ctx
dense
Phi-4 14B14B22.9 GB29 tok/s16K ctx
dense
Qwen 3 4B4B15.9 GB56 tok/s33K ctx
dense
Mistral Small 3.1 24B24B28.3 GB17 tok/s131K ctx
dense
DeepSeek Coder V2 16B16B24.3 GB57 tok/s131K ctx
moe
Llama 3.1 70B70B58.9 GB6 tok/s50K ctx
dense
Qwen 2.5 72B72B60.1 GB6 tok/s46K ctx
dense
Ministral 3 8B8B18.3 GB51 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B18.1 GB51 tok/s2K ctx
dense
Qwen 2.5 14B14B22.7 GB29 tok/s131K ctx
dense
SQLCoder 7B7B17.5 GB58 tok/s8K ctx
dense
Jina Embeddings v30.57B14.4 GB8 tok/s8K ctx
dense
Qwen 2.5 VL 7B7B16.4 GB59 tok/s33K ctx
dense
Nemotron Nano 9B v29B19.2 GB45 tok/s131K ctx
dense
Granite Code 20B20B26.6 GB21 tok/s8K ctx
dense
Gemma 3 12B12B23.5 GB25 tok/s131K ctx
dense
BGE M30.57B13.5 GB8 tok/s8K ctx
dense
Qwen 2.5 Coder 32B32B34.7 GB13 tok/s131K ctx
dense
StarCoder 15B15B36.7 GB22 tok/s8K ctx
dense
Magistral 7B7B17.5 GB58 tok/s8K ctx
dense
Granite Code 34B34B35.7 GB12 tok/s8K ctx
dense
CodeLlama 13B Instruct13B31.4 GB29 tok/s16K ctx
dense
CodeGeeX 4 9B9B17.4 GB46 tok/s131K ctx
dense
Command R 35B35B35.1 GB12 tok/s131K ctx
dense
OLMo 2 13B13B21.6 GB32 tok/s33K ctx
dense
Gemma 4 E4B8B17.4 GB39 tok/s128K ctx
dense
DeepSeek R1 Distill 32B32B34.7 GB13 tok/s33K ctx
dense
DeepSeek R1 Distill 70B70B58.9 GB6 tok/s50K ctx
dense
Qwen 2.5 7B7B16.4 GB59 tok/s131K ctx
dense
LLaVA 1.6 13B13B31.4 GB29 tok/s4K ctx
dense
DevStral 7B7B17.5 GB58 tok/s8K ctx
dense
mxbai Embed Large0.34B13.4 GB5 tok/s1K ctx
dense
CodeLlama 7B Instruct7B23.4 GB54 tok/s16K ctx
dense
MPT-30B-Instruct30B56.3 GB11 tok/s8K ctx
dense
Codestral Mamba 7B7B16.0 GB63 tok/s262K ctx
state-space
Granite Code 8B8B18.1 GB51 tok/s8K ctx
dense
StarCoder 7B7B22.9 GB54 tok/s8K ctx
dense
Snowflake Arctic Embed L0.34B13.4 GB5 tok/s1K ctx
dense
B
B Tier -- Good Fit (60 models)
Solid performance, may need quantization or reduced context.
Granite 4.1 8B8B18.6 GB51 tok/s131K ctx
dense
Qwen 2.5 Coder 3B3B15.3 GB42 tok/s131K ctx
dense
DeepSeek R1 Distill 14B14B22.7 GB29 tok/s33K ctx
dense
Nous Hermes 1.09B29.0 GB42 tok/s16K ctx
dense
Pixtral 12B12B21.0 GB34 tok/s131K ctx
dense
WizardLM 13B13B31.4 GB29 tok/s8K ctx
dense
Vicuna 13B13B31.4 GB29 tok/s4K ctx
dense
Nous Dolphin 13B13B32.8 GB25 tok/s16K ctx
dense
BGE Large EN v1.50.34B13.4 GB5 tok/s1K ctx
dense
Falcon 40B Instruct40B41.9 GB9 tok/s8K ctx
dense
Gemma 4 E2B5.1B14.9 GB62 tok/s128K ctx
dense
InternLM 7B7B23.4 GB54 tok/s8K ctx
dense
InternLM Chat 7B7B23.4 GB54 tok/s8K ctx
dense
Command A 111B111B82.9 GB3 tok/s4K ctx
dense
Nemotron 70B70B58.9 GB6 tok/s50K ctx
dense
Gemma 2 27B27B39.0 GB11 tok/s8K ctx
dense
Solar 7B7B18.5 GB58 tok/s8K ctx
dense
Llama 3.1 8B8B18.1 GB51 tok/s128K ctx
dense
Nomic Embed Text v1.50.14B12.1 GB2 tok/s8K ctx
dense
WizardMath 7B7B17.5 GB58 tok/s4K ctx
dense
OLMo 2 7B7B17.5 GB58 tok/s4K ctx
dense
GLM-4 9B9B17.4 GB46 tok/s128K ctx
dense
Gemma 3 4B4B15.8 GB56 tok/s128K ctx
dense
Ministral 3 3B3B13.8 GB42 tok/s262K ctx
multimodal
Phi 4 Mini 4B4B15.2 GB56 tok/s128K ctx
dense
Qwen 2.5 Coder 7B7B16.4 GB59 tok/s131K ctx
dense
LLaVA 1.5 7B7B23.4 GB54 tok/s4K ctx
dense
Mixtral 8x7B47B41.9 GB17 tok/s33K ctx
moe
Qwen 3.5 2B2B14.2 GB28 tok/s131K ctx
dense
Baichuan 13B13B32.8 GB25 tok/s8K ctx
dense
Llama 4 Scout 17B 16E109B80.7 GB7 tok/s4K ctx
moe
MPT-7B-Instruct7B23.4 GB54 tok/s8K ctx
dense
Baichuan 7B7B23.4 GB54 tok/s8K ctx
dense
Cerebras-GPT 13B13B30.4 GB25 tok/s79K ctx
dense
Phi 3 Mini 3.8B3.8B19.4 GB53 tok/s128K ctx
dense
Qwen 2.5 3B3B15.3 GB42 tok/s131K ctx
dense
DeepSeek R1 Distill 7B7B16.4 GB59 tok/s33K ctx
dense
DeepSeek R1 Distill 8B8B18.1 GB51 tok/s33K ctx
dense
Phi 3.5 Mini 4B4B19.6 GB56 tok/s128K ctx
dense
Samantha 7B7B17.5 GB58 tok/s4K ctx
dense
Qwen 3 1.7B1.7B14.0 GB24 tok/s33K ctx
dense
Falcon 7B Instruct7B15.7 GB60 tok/s8K ctx
dense
All MiniLM L6 v20.02B11.6 GB2 tok/s0K ctx
dense
Granite 4.1 3B3B14.3 GB42 tok/s131K ctx
dense
Granite Code 3B3B15.6 GB42 tok/s8K ctx
dense
Gemma 2 9B9B21.9 GB34 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B12.6 GB21 tok/s33K ctx
dense
Qwen 2.5 Coder 14B14B22.7 GB29 tok/s131K ctx
dense
Llama 3.2 11B Vision11B19.9 GB37 tok/s16K ctx
vision
Yi 1.5 34B34B35.7 GB12 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B60.1 GB6 tok/s4K ctx
dense
Mistral 7B Instruct v0.37B17.5 GB58 tok/s8K ctx
dense
InternLM 20B20B44.0 GB16 tok/s8K ctx
dense
Mistral Nemo 12B12B21.0 GB34 tok/s128K ctx
dense
Yi Coder 9B9B18.2 GB46 tok/s131K ctx
dense
Llama 3.2 3B3B14.8 GB42 tok/s128K ctx
dense
Phi 3 Medium 14B14B22.9 GB29 tok/s128K ctx
dense
DeepSeek LLM 67B67B57.9 GB6 tok/s4K ctx
dense
Codestral 22B22B27.1 GB19 tok/s33K ctx
dense
Ministral 8B8B18.3 GB51 tok/s131K ctx
dense
C
C Tier -- Tight Fit (193 models)
Runs with offloading or significant compromises.
SmolLM3 3B3B15.1 GB42 tok/s128K ctx
dense
DeepSeek R1 1.5B1.5B12.6 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B12.3 GB15 tok/s4K ctx
dense
Aya Expanse 32B32B33.2 GB13 tok/s8K ctx
dense
Command R+ 104B104B78.1 GB3 tok/s4K ctx
dense
Granite 3.1 8B8B18.1 GB59 tok/s128K ctx
state-space
Qwen 2.5 Math 7B7B16.4 GB59 tok/s4K ctx
dense
Yi 1.5 9B9B18.2 GB46 tok/s4K ctx
dense
Qwen 2.5 1.5B1.5B12.6 GB21 tok/s131K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4848B46.2 GB8 tok/s81K ctx
dense
Vicuna 7B7B23.4 GB54 tok/s4K ctx
dense
Yi 34B Chat34B35.7 GB12 tok/s162K ctx
dense
OpenChat 7B7B17.5 GB58 tok/s8K ctx
dense
Gemma 3 1B1B12.3 GB14 tok/s33K ctx
dense
StableLM 2 12B12B32.1 GB25 tok/s4K ctx
dense
Aya Expanse 8B8B18.1 GB51 tok/s8K ctx
dense
Gemma 2 2B2B14.1 GB28 tok/s8K ctx
dense
Qwen3.5 35B A3B35B36.7 GB11 tok/s142K ctx
dense
Qwen3.5 35B A3B
C4735B36.7 GB11 tok/s142K ctx
dense
OpenHermes 2.5 7B7B17.5 GB58 tok/s8K ctx
dense
Hermes 4.3 36B
C4636B37.4 GB11 tok/s136K ctx
dense
Zephyr 7B Beta7B17.5 GB58 tok/s33K ctx
dense
Qwen 2.5 Coder 0.5B0.5B11.8 GB7 tok/s131K ctx
dense
Dolphin 2.9 8B8B18.1 GB51 tok/s33K ctx
dense
Starling LM 7B7B17.5 GB58 tok/s8K ctx
dense
StarCoder2 15B15B23.3 GB24 tok/s16K ctx
dense
Nemotron Mini 4B4B15.7 GB56 tok/s4K ctx
dense
Qwen3.5 27B27B30.9 GB14 tok/s209K ctx
dense
EXAONE 4.0 32B
C4632B34.5 GB12 tok/s164K ctx
dense
aya expanse 32b heretic MPOA i1
C4632B34.5 GB12 tok/s164K ctx
dense
Baichuan M2 32B Q4 K M
C4632B34.5 GB12 tok/s164K ctx
dense
gemma 3 27b it27B30.9 GB14 tok/s209K ctx
dense
Yi 1.5 6B6B15.9 GB69 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
C4632B34.5 GB12 tok/s164K ctx
dense
DeepSeek LLM 7B7B22.9 GB54 tok/s4K ctx
dense
Yi 1.5 6B Chat
C466B15.6 GB63 tok/s1233K ctx
dense
Llama 2 7B Chat
C467B16.4 GB54 tok/s1045K ctx
dense
falcon mamba 7b instruct Q4 K M
C467B16.4 GB63 tok/s1045K ctx
dense
Mistral 7B Instruct v0.2
C467B16.4 GB54 tok/s1045K ctx
dense
gemma 3 27b it
C4627B30.9 GB14 tok/s209K ctx
dense
stablelm 2 zephyr 1 6b
C466B15.6 GB63 tok/s1233K ctx
dense
Mistral 7B Instruct v0.3
C457B16.4 GB54 tok/s1045K ctx
dense
Meta Llama 3.1 8B Instruct
C458B17.1 GB48 tok/s904K ctx
dense
Mamba Codestral 7B v0.1
C457B16.4 GB63 tok/s1045K ctx
dense
llava llama 3 8b v1 1
C458B17.1 GB48 tok/s904K ctx
dense
Yi 1.5 6B Chat
C456B15.6 GB63 tok/s1233K ctx
dense
stablelm 2 1 6b chat imatrix
C456B15.6 GB63 tok/s1233K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B17.1 GB48 tok/s904K ctx
dense
Qwen3.5 9B9B17.8 GB42 tok/s794K ctx
dense
Mistral Small 3.2 24B Instruct 250624B28.7 GB16 tok/s246K ctx
dense
Mistral 7B Instruct v0.3
C457B16.4 GB54 tok/s1045K ctx
dense
Meta Llama 3 8B Instruct
C458B17.1 GB48 tok/s904K ctx
dense
Hermes 2 Pro Mistral 7B
C457B16.4 GB54 tok/s1045K ctx
dense
DeepSeek R1 Distill Llama 8B8B17.1 GB48 tok/s904K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C459B17.8 GB42 tok/s794K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B17.1 GB48 tok/s904K ctx
dense
Llama 3 8B Instruct 32k v0.1
C458B17.1 GB48 tok/s904K ctx
dense
zephyr 7B beta
C457B16.4 GB54 tok/s1045K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4524B28.7 GB16 tok/s246K ctx
dense
Meta Llama 3.1 8B Instruct
C458B17.1 GB48 tok/s904K ctx
dense
vntl llama3 8b v2
C458B17.1 GB48 tok/s904K ctx
dense
Mistral Small 24B Instruct 2501
C4524B28.7 GB16 tok/s246K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C457B16.4 GB54 tok/s1045K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B17.1 GB48 tok/s904K ctx
dense
Falcon H1R 7B
C457B16.4 GB54 tok/s1045K ctx
dense
HelpingAI2.5 5B i1
C455B14.9 GB70 tok/s1496K ctx
dense
Llama 3.3 70B Instruct
C4570B62.2 GB5 tok/s30K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C458B17.1 GB48 tok/s904K ctx
dense
Qwen3.5 9B
C459B17.8 GB42 tok/s794K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4524B28.7 GB16 tok/s246K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4524B28.7 GB16 tok/s246K ctx
dense
Neural Chat 7B7B17.5 GB58 tok/s8K ctx
dense
zephyr 7B alpha
C457B16.4 GB54 tok/s1045K ctx
dense
speechless zephyr code functionary 7b
C457B16.4 GB54 tok/s1045K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C458B17.1 GB48 tok/s904K ctx
dense
StarCoder2 7B
C457B16.4 GB54 tok/s1045K ctx
dense
HelpingAI2 6B
C456B15.6 GB63 tok/s1233K ctx
dense
Hermes 3 Llama 3.1 8B
C458B17.1 GB48 tok/s904K ctx
dense
japanese stablelm instruct gamma 7B
C457B16.4 GB54 tok/s1045K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4524B28.7 GB16 tok/s246K ctx
dense
HelpingAI2 6B i1
C456B15.6 GB63 tok/s1233K ctx
dense
Falcon H1 7B Instruct
C457B16.4 GB54 tok/s1045K ctx
dense
Falcon H1R 7B7B16.4 GB54 tok/s1045K ctx
dense
Qwen3.5 4B4B14.2 GB56 tok/s1891K ctx
dense
Hermes 2 Pro Llama 3 8B
C458B17.1 GB48 tok/s904K ctx
dense
Dolphin3.0 Llama3.1 8B
C458B17.1 GB48 tok/s904K ctx
dense
granite 8b code instruct 4k
C458B17.1 GB48 tok/s904K ctx
dense
Yi Coder 9B Chat
C459B17.8 GB42 tok/s794K ctx
dense
dolphin 2.9.4 llama3.1 8b
C458B17.1 GB48 tok/s904K ctx
dense
glm 4 9b chat 1m
C459B17.8 GB42 tok/s794K ctx
dense
Codestral 22B v0.1
C4522B27.3 GB17 tok/s276K ctx
dense
gemma 3 4b it
C454B14.2 GB56 tok/s1891K ctx
dense
Codestral 22B v0.1
C4522B27.3 GB17 tok/s276K ctx
dense
stabilityai japanese stablelm base gamma 7b
C457B16.4 GB54 tok/s1045K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B16.9 GB49 tok/s929K ctx
dense
baichuan2 7b chat
C457B16.4 GB54 tok/s1045K ctx
dense
openchat 3.6 8b 20240522 IMat
C458B17.1 GB48 tok/s904K ctx
dense
DiscoPOP zephyr 7b gemma
C457B16.4 GB54 tok/s1045K ctx
dense
aya expanse 8b
C458B17.1 GB48 tok/s904K ctx
dense
EXAONE 3.5 7.8B Instruct
C457.8B16.9 GB49 tok/s929K ctx
dense
starcoder2 7b
C457B16.4 GB54 tok/s1045K ctx
dense
Solar Open 69B REAP i1
C4569B61.4 GB6 tok/s31K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C457B16.4 GB54 tok/s1045K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C457B16.4 GB54 tok/s1045K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C457.8B16.9 GB49 tok/s929K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C457B16.4 GB54 tok/s1045K ctx
dense
dolphin v2 8b abliterated i1
C458B17.1 GB48 tok/s904K ctx
dense
internlm2 math plus 7b IMat
C457B16.4 GB54 tok/s1045K ctx
dense
gemma 3 4b it
C454B14.2 GB56 tok/s1891K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C457B16.4 GB54 tok/s1045K ctx
dense
Yi 1.5 9B Chat
C459B17.8 GB42 tok/s794K ctx
dense
exaone 3.0 7.8b it
C457.8B16.9 GB49 tok/s929K ctx
dense
internlm2 5 7b chat i1
C457B16.4 GB54 tok/s1045K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4524B28.7 GB16 tok/s246K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C457B16.4 GB54 tok/s1045K ctx
dense
MD Judge v0 2 internlm2 7b i1
C457B16.4 GB54 tok/s1045K ctx
dense
aya expanse 8b orthogonal heretic i1
C458B17.1 GB48 tok/s904K ctx
dense
gemma 3 12b it
C4512B20.0 GB32 tok/s575K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C457B16.4 GB54 tok/s1045K ctx
dense
Codestral 22B v0.1
C4522B27.3 GB17 tok/s276K ctx
dense
jointpreferences mistral 7b sft helpful
C457B16.4 GB54 tok/s1045K ctx
dense
zephyr 7b dpo full i1
C457B16.4 GB54 tok/s1045K ctx
dense
blossom v3 baichuan2 7b i1
C457B16.4 GB54 tok/s1045K ctx
dense
DeepSeek R1 Distill Qwen 14B14B21.4 GB27 tok/s481K ctx
dense
blossom v1 baichuan 7b i1
C457B16.4 GB54 tok/s1045K ctx
dense
aya expanse 8b orthogonal heretic
C458B17.1 GB48 tok/s904K ctx
dense
StarCoder2 7B7B16.0 GB59 tok/s16K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4510.7B19.0 GB36 tok/s655K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C4570B62.2 GB5 tok/s30K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4510.7B19.0 GB36 tok/s655K ctx
dense
Codestral 22B v0.1 IMat
C4522B27.3 GB17 tok/s276K ctx
dense
internlm2 5 1 8b chat i1
C458B17.1 GB48 tok/s904K ctx
dense
internlm3 8b instruct abliterated i1
C458B17.1 GB48 tok/s904K ctx
dense
Yi 9B Coder i1
C459B17.8 GB42 tok/s794K ctx
dense
Phi 4 reasoning vision 15B
C4515B22.2 GB25 tok/s443K ctx
dense
StarCoder2 15B
C4515B22.2 GB25 tok/s443K ctx
dense
HelpingAI2 9B
C449B17.8 GB42 tok/s794K ctx
dense
NousResearch Hermes 4 14B
C4414B21.4 GB27 tok/s481K ctx
dense
Codestral 22B v0.1 i1
C4422B27.3 GB17 tok/s276K ctx
dense
internlm2 5 20b chat
C4420B25.8 GB19 tok/s312K ctx
dense
Qwen 3 0.6B0.6B12.5 GB8 tok/s33K ctx
dense
BaichuanMed OCR 72B i1
C4472B63.6 GB5 tok/s26K ctx
dense
HelpingAI2 9B i1
C449B17.8 GB42 tok/s794K ctx
dense
HelpingAI 9B 200k i1
C449B17.8 GB42 tok/s794K ctx
dense
Codestral 21B Pruned i1
C4421B26.5 GB18 tok/s293K ctx
dense
internlm2 limarp chat 20b
C4420B25.8 GB19 tok/s312K ctx
dense
HelpingAI 9B i1
C449B17.8 GB42 tok/s794K ctx
dense
solar finalised finetuned Model 10.7B i1
C4410.7B19.0 GB36 tok/s655K ctx
dense
internlm2 math plus 20b i1
C4420B25.8 GB19 tok/s312K ctx
dense
Llama 3.2 1B1B12.4 GB14 tok/s128K ctx
dense
SOLAR 10.7B v1.0
C4410.7B19.0 GB36 tok/s655K ctx
dense
starcoder2 15b instruct v0.1
C4415B22.2 GB25 tok/s443K ctx
dense
starcoder2 15b i1
C4415B22.2 GB25 tok/s443K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4414B21.4 GB27 tok/s481K ctx
dense
HelpingAI2.5 10B i1
C4410B18.5 GB38 tok/s707K ctx
dense
starcoder2 15b instruct v0.1
C4415B22.2 GB25 tok/s443K ctx
dense
Helply 10.2b chat i1
C4410.2B18.7 GB37 tok/s691K ctx
dense
Codestral RAG 19B Pruned i1
C4419B25.1 GB20 tok/s332K ctx
dense
internlm JanusCoder 14B
C4414B21.4 GB27 tok/s481K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4414B21.4 GB27 tok/s481K ctx
dense
HelpingAI 15B i1
C4415B22.2 GB25 tok/s443K ctx
dense
Llama 3.2 3B Instruct
C443B13.8 GB42 tok/s2535K ctx
dense
Qwen2.5 3B Instruct3B13.4 GB42 tok/s2550K ctx
dense
Llama 3.2 3B Instruct
C443B13.4 GB42 tok/s2550K ctx
dense
Ministral 3 3B Instruct 2512
C443B13.4 GB42 tok/s2550K ctx
dense
HELVETE 3B
C433B13.4 GB42 tok/s2550K ctx
dense
Hermes 3 Llama 3.2 3B
C433B13.4 GB42 tok/s2550K ctx
dense
stablelm zephyr 3b
C433B13.4 GB42 tok/s2550K ctx
dense
StarCoder2 3B
C433B13.4 GB42 tok/s2550K ctx
dense
AI21 Jamba Reasoning 3B
C433B13.4 GB42 tok/s2550K ctx
dense
stablelm 3b 4e1t
C433B13.4 GB42 tok/s2550K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C433B13.4 GB42 tok/s2550K ctx
dense
ai21labs AI21 Jamba2 3B
C433B13.4 GB42 tok/s2550K ctx
dense
HelpingAI 3B hindi i1
C433B13.4 GB42 tok/s2550K ctx
dense
AI21 Jamba2 3B
C433B13.4 GB42 tok/s2550K ctx
dense
HelpingAI 3B hindi
C433B13.4 GB42 tok/s2550K ctx
dense
AI21 Jamba2 3B i1
C433B13.4 GB42 tok/s2550K ctx
dense
EXAONE 3.5 2.4B Instruct
C422.4B13.0 GB34 tok/s3208K ctx
dense
gemma 2 2b it
C422B13.1 GB28 tok/s3837K ctx
dense
gemma 2b2B12.7 GB28 tok/s3866K ctx
dense
Gemmasutra Mini 2B v1
C422B12.7 GB28 tok/s3866K ctx
dense
StarCoder2 3B3B13.6 GB42 tok/s16K ctx
dense
gemma 2 2b it
C422B12.7 GB28 tok/s3866K ctx
dense
Qwen 3.5 0.6B0.6B12.5 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B12.4 GB21 tok/s5183K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B12.4 GB21 tok/s5183K ctx
dense
Yi Coder 1.5B Chat
C411.5B12.4 GB21 tok/s5183K ctx
dense
stablelm 2 zephyr 1.6b
C411.6B12.4 GB22 tok/s4853K ctx
dense
logos16v2 stablelm2 1.6b i1
C411.6B12.4 GB22 tok/s4853K ctx
dense
Yi Coder 1.5B
C411.5B12.4 GB21 tok/s5183K ctx
dense
Falcon H1 1.5B Instruct1.5B12.4 GB21 tok/s5183K ctx
dense
TinyLlama 1.1B Chat v1.0
C401.1B12.1 GB15 tok/s7097K ctx
dense
Qwen 2.5 0.5B0.5B11.8 GB7 tok/s131K ctx
dense
F
F Tier -- Can't Run (45 models)
Too heavy for this hardware.
EXAONE 4.0 1.2B
D401.2B12.1 GB17 tok/s6499K ctx
dense
Llama 3.2 1B Instruct Q8 0
D401B12.2 GB14 tok/s7787K ctx
dense
TinyLlama 1.1B Chat v0.3
D401.1B12.1 GB15 tok/s7097K ctx
dense
TinyLlama 1.1B Chat v0.61.1B12.1 GB15 tok/s7097K ctx
dense
Llama 3.2 1B Instruct
D401B12.0 GB14 tok/s7815K ctx
dense
gemma 3 1b it
D401B12.0 GB14 tok/s7815K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D391.1B12.1 GB15 tok/s7097K ctx
dense
Falcon3 1B Instruct abliterated
D391B12.0 GB14 tok/s7815K ctx
dense
SmolVLM 500M Instruct
D380.5B11.8 GB7 tok/s9191K ctx
dense
embeddinggemma 300M
D360.3B11.6 GB4 tok/s9217K ctx
dense
granite embedding 107m multilingual
D340.11B11.4 GB2 tok/s9246K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B11.4 GB2 tok/s9248K ctx
dense
Qwen 3 235B A22B235B157.5 GB2 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B306.9 GB2 tok/s4K ctx
moe
Qwen 3.5 122B A10B122B88.1 GB10 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B256.3 GB2 tok/s4K ctx
moe
Mixtral 8x22B141B100.7 GB3 tok/s4K ctx
moe
Devstral 2 123B Instruct123B91.7 GB2 tok/s4K ctx
dense
Mistral Small 4 119B119B89.2 GB11 tok/s4K ctx
moe
Leanstral 119B A6B119B92.6 GB10 tok/s4K ctx
moe
Mistral Large 3675B428.4 GB2 tok/s4K ctx
moe
GLM-5744B484.2 GB2 tok/s4K ctx
moe
GLM-5.1754B490.3 GB2 tok/s4K ctx
moe
Kimi K2.51000B628.7 GB2 tok/s4K ctx
moe
Kimi K2.61000B628.7 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B480.1 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B480.1 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B875.1 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B170.6 GB4 tok/s4K ctx
moe
DeepSeek R1 671B671B480.1 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B213.8 GB3 tok/s4K ctx
moe
DeepSeek V2.5 236B236B213.8 GB3 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B258.2 GB2 tok/s4K ctx
moe
Pixtral Large 124B124B92.3 GB2 tok/s4K ctx
dense
Llama 3.1 405B405B266.0 GB2 tok/s4K ctx
dense
GPT-OSS 120B117B87.5 GB3 tok/s4K ctx
dense
DeepSeek V3.2671B421.0 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B155.4 GB4 tok/s4K ctx
moe
Solar Open 100B
F0100B84.0 GB3 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B182.9 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B182.2 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B182.2 GB2 tok/s4K ctx
dense
Solar Open 100B i1
F0100B84.0 GB3 tok/s4K ctx
dense
Qwen3.5 122B A10B122B85.3 GB3 tok/s4K ctx
dense
Qwen3.5 397B A17B397B300.0 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs