AI Compatibility Tier List
MacBook Pro M4 Max 48GB
380 AI models ranked by how well they run on your 48 GB unified memory hardware. Best match: Qwen3-Coder 30B A3B Instruct (score: 98).
380
Models tested
322
Can run
48 GB
unified memory
52
Best tok/s
Tier Distribution
S 27
A 62
B 38
C 195
F 58
S: 27A: 62B: 38C: 195F: 58
S
S Tier -- Perfect Fit (27 models)
These models run flawlessly with headroom to spare.
Qwen3-Coder 30B A3B Instruct30.5B26.2 GB52 tok/s108K ctx
moe
Qwen3-VL 30B A3B Instruct30B25.8 GB54 tok/s111K ctx
moe
Qwen 3.5 27B27B25.7 GB36 tok/s61K ctx
dense
Qwen 3 30B A3B30.5B26.2 GB52 tok/s108K ctx
moe
Qwen 3.6 27B27B23.5 GB27 tok/s197K ctx
dense
Magistral Small 250724B23.2 GB37 tok/s91K ctx
dense
Devstral Small 2 24B Instruct24B23.2 GB37 tok/s91K ctx
dense
Nemotron 3 Nano 30B30B26.8 GB38 tok/s67K ctx
dense
Qwen 3.6 35B A3B35B31.5 GB44 tok/s28K ctx
moe
Qwen 3.5 35B A3B35B28.9 GB48 tok/s78K ctx
moe
Nemotron Cascade 2 30B A3B30B27.3 GB53 tok/s56K ctx
moe
Devstral Small 1.124B23.2 GB37 tok/s91K ctx
dense
GPT-OSS 20B21B21.3 GB66 tok/s103K ctx
moe
Qwen 3 32B32B29.5 GB34 tok/s37K ctx
dense
Qwen 3.5 9B9B13.8 GB73 tok/s131K ctx
dense
Qwen 3 14B14B17.1 GB38 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B25.1 GB56 tok/s57K ctx
moe
Phi-4-reasoning-plus 14B14.7B18.1 GB38 tok/s33K ctx
dense
Qwen 3 8B8B13.2 GB83 tok/s131K ctx
dense
Mistral Small 3.2 24B24B23.2 GB37 tok/s91K ctx
vision
Qwen 3.5 4B4B10.7 GB56 tok/s131K ctx
dense
Codestral 2 25.0822B21.9 GB35 tok/s99K ctx
dense
LFM2 24B24B23.2 GB37 tok/s91K ctx
dense
Granite 4.1 30B30B28.3 GB35 tok/s42K ctx
dense
EXAONE 4.0 32B32B29.5 GB33 tok/s37K ctx
dense
Ornith 1.0 35B A3B
S8535.1B27.8 GB47 tok/s262K ctx
moe
Mistral Small 24B24B23.2 GB37 tok/s33K ctx
dense
A
A Tier -- Great Fit (62 models)
Excellent performance with comfortable memory margins.
Nemotron Nano 8B8B12.9 GB89 tok/s131K ctx
dense
CogVLM2 19B19B20.1 GB39 tok/s8K ctx
dense
Ministral 3 14B14B17.1 GB38 tok/s131K ctx
multimodal
Mistral Small 3.1 24B24B23.2 GB37 tok/s91K ctx
dense
Qwen 2.5 32B32B29.5 GB33 tok/s37K ctx
dense
DiffusionGemma 26B A4B25.8B25.5 GB54 tok/s56K ctx
moe
Phi-4 Mini Reasoning 4B3.8B9.9 GB53 tok/s131K ctx
dense
OLMo 2 32B32B29.5 GB33 tok/s4K ctx
dense
Gemma 3 27B27B33.8 GB27 tok/s17K ctx
dense
DeepSeek Coder V2 16B16B19.1 GB84 tok/s91K ctx
moe
Granite Code 20B20B21.5 GB38 tok/s8K ctx
dense
Gemma 4 12B12B19.3 GB33 tok/s58K ctx
dense
InternVL2 8B8B12.9 GB83 tok/s8K ctx
dense
Phi-4 14B14B17.7 GB38 tok/s16K ctx
dense
Kimi Linear 48B A3B48B36.3 GB19 tok/s4K ctx
linear
Qwen 2.5 14B14B17.6 GB38 tok/s109K ctx
dense
Ministral 3 8B8B13.2 GB83 tok/s172K ctx
multimodal
MiniCPM-V 2.6 8B8B12.9 GB83 tok/s2K ctx
dense
Gemma 3 12B12B18.3 GB33 tok/s69K ctx
dense
Nemotron Nano 9B v29B14.0 GB79 tok/s131K ctx
dense
CodeLlama 13B Instruct13B26.2 GB38 tok/s16K ctx
dense
SQLCoder 7B7B12.3 GB94 tok/s8K ctx
dense
Qwen 2.5 VL 7B7B11.2 GB95 tok/s33K ctx
dense
Qwen 2.5 Coder 32B32B29.5 GB33 tok/s37K ctx
dense
Qwen 3 4B4B10.7 GB56 tok/s33K ctx
dense
LLaVA 1.6 13B13B26.2 GB38 tok/s4K ctx
dense
Granite Code 34B34B30.5 GB31 tok/s8K ctx
dense
Magistral 7B7B12.3 GB94 tok/s8K ctx
dense
Jina Embeddings v30.57B9.2 GB8 tok/s8K ctx
dense
Command R 35B35B29.9 GB31 tok/s47K ctx
dense
CodeGeeX 4 9B9B12.2 GB75 tok/s131K ctx
dense
OLMo 2 13B13B16.5 GB41 tok/s33K ctx
dense
CodeLlama 7B Instruct7B18.2 GB88 tok/s16K ctx
dense
Nous Hermes 1.09B23.8 GB68 tok/s16K ctx
dense
StarCoder 7B7B17.7 GB88 tok/s8K ctx
dense
DeepSeek R1 Distill 32B32B29.5 GB33 tok/s33K ctx
dense
Gemma 4 31B30.7B39.5 GB21 tok/s11K ctx
dense
Gemma 4 E4B8B12.2 GB63 tok/s128K ctx
dense
StarCoder 15B15B31.5 GB30 tok/s8K ctx
dense
BGE M30.57B8.3 GB8 tok/s8K ctx
dense
Qwen 2.5 7B7B11.2 GB95 tok/s131K ctx
dense
Granite Code 8B8B12.9 GB83 tok/s8K ctx
dense
WizardLM 13B13B26.2 GB38 tok/s8K ctx
dense
Vicuna 13B13B26.2 GB38 tok/s4K ctx
dense
DevStral 7B7B12.3 GB94 tok/s8K ctx
dense
DeepSeek R1 Distill 14B14B17.6 GB38 tok/s33K ctx
dense
Nous Dolphin 13B13B27.7 GB33 tok/s16K ctx
dense
Granite 4.1 8B8B13.4 GB83 tok/s131K ctx
dense
Codestral Mamba 7B7B10.8 GB93 tok/s262K ctx
state-space
InternLM 7B7B18.2 GB88 tok/s8K ctx
dense
InternLM Chat 7B7B18.2 GB88 tok/s8K ctx
dense
LFM2.5 8B A1B8.5B11.5 GB162 tok/s128K ctx
moe
Pixtral 12B12B15.8 GB45 tok/s131K ctx
dense
mxbai Embed Large0.34B8.2 GB5 tok/s1K ctx
dense
Qwen 2.5 Coder 3B3B10.1 GB42 tok/s131K ctx
dense
LLaVA 1.5 7B7B18.2 GB88 tok/s4K ctx
dense
Snowflake Arctic Embed L0.34B8.2 GB5 tok/s1K ctx
dense
Solar 7B7B13.3 GB94 tok/s8K ctx
dense
Llama 3.1 8B8B12.9 GB83 tok/s128K ctx
dense
GLM-4 9B9B12.2 GB75 tok/s128K ctx
dense
WizardMath 7B7B12.3 GB94 tok/s4K ctx
dense
OLMo 2 7B7B12.3 GB94 tok/s4K ctx
dense
B
B Tier -- Good Fit (38 models)
Solid performance, may need quantization or reduced context.
Gemma 4 E2B5.1B9.7 GB71 tok/s128K ctx
dense
Baichuan 13B13B27.7 GB33 tok/s8K ctx
dense
Cerebras-GPT 13B13B25.2 GB33 tok/s31K ctx
dense
Gemma 2 27B27B33.8 GB27 tok/s8K ctx
dense
MPT-7B-Instruct7B18.2 GB88 tok/s8K ctx
dense
Baichuan 7B7B18.2 GB88 tok/s8K ctx
dense
BGE Large EN v1.50.34B8.2 GB5 tok/s1K ctx
dense
Qwen 2.5 Coder 7B7B11.2 GB95 tok/s131K ctx
dense
Gemma 3 4B4B10.6 GB56 tok/s128K ctx
dense
Nomic Embed Text v1.50.14B6.9 GB2 tok/s8K ctx
dense
Ministral 3 3B3B8.6 GB42 tok/s262K ctx
multimodal
Phi 4 Mini 4B4B10.0 GB56 tok/s128K ctx
dense
DeepSeek R1 Distill 8B8B12.9 GB83 tok/s33K ctx
dense
Phi 3 Mini 3.8B3.8B14.3 GB53 tok/s71K ctx
dense
Qwen 3.5 2B2B9.0 GB28 tok/s131K ctx
dense
DeepSeek R1 Distill 7B7B11.2 GB95 tok/s33K ctx
dense
Samantha 7B7B12.3 GB94 tok/s4K ctx
dense
Gemma 2 9B9B16.7 GB54 tok/s8K ctx
dense
Phi 3.5 Mini 4B4B14.4 GB56 tok/s71K ctx
dense
Falcon 7B Instruct7B10.5 GB97 tok/s8K ctx
dense
Qwen 2.5 3B3B10.1 GB42 tok/s131K ctx
dense
Qwen 2.5 Coder 14B14B17.6 GB38 tok/s109K ctx
dense
Llama 3.2 11B Vision11B14.7 GB49 tok/s16K ctx
vision
Yi 1.5 34B34B30.5 GB31 tok/s4K ctx
dense
Qwen 3 1.7B1.7B8.8 GB24 tok/s33K ctx
dense
Codestral 22B22B21.9 GB37 tok/s33K ctx
dense
Granite 4.1 3B3B9.1 GB42 tok/s131K ctx
dense
Granite Code 3B3B10.4 GB42 tok/s8K ctx
dense
Mistral 7B Instruct v0.37B12.3 GB94 tok/s8K ctx
dense
Mistral Nemo 12B12B15.8 GB45 tok/s128K ctx
dense
Yi Coder 9B9B13.0 GB74 tok/s131K ctx
dense
All MiniLM L6 v20.02B6.4 GB2 tok/s0K ctx
dense
Phi 3 Medium 14B14B17.7 GB38 tok/s105K ctx
dense
Qwen 2.5 Coder 1.5B1.5B7.4 GB21 tok/s33K ctx
dense
Falcon 40B Instruct40B36.7 GB21 tok/s4K ctx
dense
Ministral 8B8B13.2 GB83 tok/s131K ctx
dense
Aya Expanse 32B32B28.0 GB34 tok/s8K ctx
dense
Llama 3.2 3B3B9.6 GB42 tok/s128K ctx
dense
C
C Tier -- Tight Fit (195 models)
Runs with offloading or significant compromises.
Mixtral 8x7B47B36.7 GB22 tok/s4K ctx
moe
Granite 3.1 8B8B12.9 GB87 tok/s128K ctx
state-space
Qwen3.5 27B27B25.7 GB33 tok/s61K ctx
dense
gemma 3 27b it27B25.7 GB33 tok/s61K ctx
dense
Yi 1.5 9B9B13.0 GB74 tok/s4K ctx
dense
gemma 3 27b it
C5427B25.7 GB33 tok/s61K ctx
dense
Vicuna 7B7B18.2 GB88 tok/s4K ctx
dense
Mistral Small 3.2 24B Instruct 250624B23.5 GB34 tok/s79K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C5324B23.5 GB34 tok/s79K ctx
dense
Mistral Small 24B Instruct 2501
C5324B23.5 GB34 tok/s79K ctx
dense
StableLM 2 12B12B26.9 GB33 tok/s4K ctx
dense
SmolLM3 3B3B9.9 GB42 tok/s128K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C5324B23.5 GB34 tok/s79K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C5324B23.5 GB34 tok/s79K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C5324B23.5 GB34 tok/s79K ctx
dense
Qwen 2.5 Math 7B7B11.2 GB95 tok/s4K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C5324B23.5 GB34 tok/s79K ctx
dense
DeepSeek R1 1.5B1.5B7.4 GB21 tok/s33K ctx
dense
Codestral 22B v0.1
C5222B22.1 GB35 tok/s93K ctx
dense
Codestral 22B v0.1
C5222B22.1 GB35 tok/s93K ctx
dense
Codestral 22B v0.1
C5222B22.1 GB35 tok/s93K ctx
dense
TinyLlama 1.1B1.1B7.1 GB15 tok/s4K ctx
dense
Codestral 22B v0.1 IMat
C5222B22.1 GB35 tok/s93K ctx
dense
Yi 34B Chat34B30.5 GB31 tok/s34K ctx
dense
Codestral 22B v0.1 i1
C5222B22.1 GB35 tok/s93K ctx
dense
Codestral 21B Pruned i1
C5221B21.4 GB35 tok/s102K ctx
dense
OpenChat 7B7B12.3 GB94 tok/s8K ctx
dense
internlm2 5 20b chat
C5120B20.6 GB36 tok/s111K ctx
dense
internlm2 limarp chat 20b
C5120B20.6 GB36 tok/s111K ctx
dense
StarCoder2 15B15B18.1 GB33 tok/s16K ctx
dense
Aya Expanse 8B8B12.9 GB83 tok/s8K ctx
dense
internlm2 math plus 20b i1
C5120B20.6 GB36 tok/s111K ctx
dense
DeepSeek LLM 7B7B17.7 GB88 tok/s4K ctx
dense
Codestral RAG 19B Pruned i1
C5119B19.9 GB36 tok/s121K ctx
dense
Qwen3.5 35B A3B35B31.5 GB28 tok/s28K ctx
dense
OpenHermes 2.5 7B7B12.3 GB94 tok/s8K ctx
dense
Qwen3.5 35B A3B
C5035B31.5 GB28 tok/s28K ctx
dense
Dolphin 2.9 8B8B12.9 GB83 tok/s33K ctx
dense
EXAONE 4.0 32B
C5032B29.4 GB31 tok/s38K ctx
dense
aya expanse 32b heretic MPOA i1
C5032B29.4 GB31 tok/s38K ctx
dense
Zephyr 7B Beta7B12.3 GB94 tok/s33K ctx
dense
Baichuan M2 32B Q4 K M
C5032B29.4 GB31 tok/s38K ctx
dense
baichuan inc Baichuan M2 32B
C5032B29.4 GB31 tok/s38K ctx
dense
Qwen 2.5 1.5B1.5B7.4 GB21 tok/s131K ctx
dense
Hermes 4.3 36B
C5036B32.3 GB27 tok/s25K ctx
dense
Starling LM 7B7B12.3 GB94 tok/s8K ctx
dense
Phi 4 reasoning vision 15B
C4915B17.0 GB35 tok/s176K ctx
dense
StarCoder2 15B
C4915B17.0 GB35 tok/s176K ctx
dense
Qwen3.5 9B9B12.6 GB68 tok/s349K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C499B12.6 GB68 tok/s349K ctx
dense
Meta Llama 3.1 8B Instruct
C498B11.9 GB77 tok/s403K ctx
dense
DeepSeek R1 Distill Qwen 14B14B16.3 GB35 tok/s194K ctx
dense
Llama 2 7B Chat
C497B11.2 GB88 tok/s472K ctx
dense
llava llama 3 8b v1 1
C498B11.9 GB77 tok/s403K ctx
dense
Qwen3.5 9B
C499B12.6 GB68 tok/s349K ctx
dense
Mistral 7B Instruct v0.2
C497B11.2 GB88 tok/s472K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B11.9 GB77 tok/s403K ctx
dense
starcoder2 15b instruct v0.1
C4915B17.0 GB35 tok/s176K ctx
dense
starcoder2 15b i1
C4915B17.0 GB35 tok/s176K ctx
dense
Mistral 7B Instruct v0.3
C497B11.2 GB88 tok/s472K ctx
dense
Meta Llama 3 8B Instruct
C498B11.9 GB77 tok/s403K ctx
dense
starcoder2 15b instruct v0.1
C4915B17.0 GB35 tok/s176K ctx
dense
DeepSeek R1 Distill Llama 8B8B11.9 GB77 tok/s403K ctx
dense
Dolphin3.0 Llama3.1 8B
C498B11.9 GB77 tok/s403K ctx
dense
Llama 3 8B Instruct 32k v0.1
C498B11.9 GB77 tok/s403K ctx
dense
Meta Llama 3.1 8B Instruct
C498B11.9 GB77 tok/s403K ctx
dense
vntl llama3 8b v2
C498B11.9 GB77 tok/s403K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C498B11.9 GB77 tok/s403K ctx
dense
HelpingAI 15B i1
C4915B17.0 GB35 tok/s176K ctx
dense
Yi Coder 9B Chat
C499B12.6 GB68 tok/s349K ctx
dense
glm 4 9b chat 1m
C499B12.6 GB68 tok/s349K ctx
dense
NousResearch Hermes 4 14B
C4914B16.3 GB35 tok/s194K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C498B11.9 GB77 tok/s403K ctx
dense
Neural Chat 7B7B12.3 GB94 tok/s8K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C498B11.9 GB77 tok/s403K ctx
dense
Hermes 3 Llama 3.1 8B
C498B11.9 GB77 tok/s403K ctx
dense
Mistral 7B Instruct v0.3
C497B11.2 GB88 tok/s472K ctx
dense
Hermes 2 Pro Mistral 7B
C497B11.2 GB88 tok/s472K ctx
dense
falcon mamba 7b instruct Q4 K M
C497B11.2 GB93 tok/s472K ctx
dense
zephyr 7B beta
C497B11.2 GB88 tok/s472K ctx
dense
Yi 1.5 9B Chat
C499B12.6 GB68 tok/s349K ctx
dense
Hermes 2 Pro Llama 3 8B
C498B11.9 GB77 tok/s403K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C497B11.2 GB88 tok/s472K ctx
dense
Dolphin3.0 Llama3.1 8B
C498B11.9 GB77 tok/s403K ctx
dense
granite 8b code instruct 4k
C498B11.9 GB77 tok/s403K ctx
dense
Falcon H1R 7B
C497B11.2 GB88 tok/s472K ctx
dense
dolphin 2.9.4 llama3.1 8b
C498B11.9 GB77 tok/s403K ctx
dense
gemma 3 12b it
C4912B14.8 GB41 tok/s241K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4914B16.3 GB35 tok/s194K ctx
dense
zephyr 7B alpha
C497B11.2 GB88 tok/s472K ctx
dense
speechless zephyr code functionary 7b
C497B11.2 GB88 tok/s472K ctx
dense
EXAONE 3.5 7.8B Instruct
C497.8B11.8 GB79 tok/s415K ctx
dense
internlm JanusCoder 14B
C4914B16.3 GB35 tok/s194K ctx
dense
openchat 3.6 8b 20240522 IMat
C498B11.9 GB77 tok/s403K ctx
dense
StarCoder2 7B
C487B11.2 GB88 tok/s472K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4814B16.3 GB35 tok/s194K ctx
dense
aya expanse 8b
C488B11.9 GB77 tok/s403K ctx
dense
EXAONE 3.5 7.8B Instruct
C487.8B11.8 GB79 tok/s415K ctx
dense
japanese stablelm instruct gamma 7B
C487B11.2 GB88 tok/s472K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C487.8B11.8 GB79 tok/s415K ctx
dense
dolphin v2 8b abliterated i1
C488B11.9 GB77 tok/s403K ctx
dense
Mamba Codestral 7B v0.1
C487B11.2 GB93 tok/s472K ctx
dense
Falcon H1 7B Instruct
C487B11.2 GB88 tok/s472K ctx
dense
Falcon H1R 7B7B11.2 GB88 tok/s472K ctx
dense
LFM2.5 350M0.35B6.5 GB5 tok/s128K ctx
dense
exaone 3.0 7.8b it
C487.8B11.8 GB79 tok/s415K ctx
dense
Yi 9B Coder i1
C489B12.6 GB68 tok/s349K ctx
dense
aya expanse 8b orthogonal heretic i1
C488B11.9 GB77 tok/s403K ctx
dense
HelpingAI2.5 10B i1
C4810B13.4 GB62 tok/s306K ctx
dense
HelpingAI2 9B
C489B12.6 GB68 tok/s349K ctx
dense
stabilityai japanese stablelm base gamma 7b
C487B11.2 GB88 tok/s472K ctx
dense
baichuan2 7b chat
C487B11.2 GB88 tok/s472K ctx
dense
aya expanse 8b orthogonal heretic
C488B11.9 GB77 tok/s403K ctx
dense
DiscoPOP zephyr 7b gemma
C487B11.2 GB88 tok/s472K ctx
dense
Gemma 2 2B2B8.9 GB28 tok/s8K ctx
dense
HelpingAI2 9B i1
C489B12.6 GB68 tok/s349K ctx
dense
starcoder2 7b
C487B11.2 GB88 tok/s472K ctx
dense
HelpingAI 9B 200k i1
C489B12.6 GB68 tok/s349K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C487B11.2 GB88 tok/s472K ctx
dense
Gemma 3 1B1B7.1 GB14 tok/s33K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C487B11.2 GB88 tok/s472K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C487B11.2 GB88 tok/s472K ctx
dense
internlm2 math plus 7b IMat
C487B11.2 GB88 tok/s472K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4810.7B13.9 GB46 tok/s280K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C487B11.2 GB88 tok/s472K ctx
dense
HelpingAI 9B i1
C489B12.6 GB68 tok/s349K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4810.7B13.9 GB46 tok/s280K ctx
dense
internlm2 5 1 8b chat i1
C488B11.9 GB77 tok/s403K ctx
dense
internlm2 5 7b chat i1
C487B11.2 GB88 tok/s472K ctx
dense
internlm3 8b instruct abliterated i1
C488B11.9 GB77 tok/s403K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C487B11.2 GB88 tok/s472K ctx
dense
MD Judge v0 2 internlm2 7b i1
C487B11.2 GB88 tok/s472K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C487B11.2 GB88 tok/s472K ctx
dense
jointpreferences mistral 7b sft helpful
C487B11.2 GB88 tok/s472K ctx
dense
zephyr 7b dpo full i1
C487B11.2 GB88 tok/s472K ctx
dense
blossom v3 baichuan2 7b i1
C487B11.2 GB88 tok/s472K ctx
dense
blossom v1 baichuan 7b i1
C487B11.2 GB88 tok/s472K ctx
dense
Yi 1.5 6B Chat
C486B10.4 GB84 tok/s565K ctx
dense
stablelm 2 zephyr 1 6b
C486B10.4 GB84 tok/s565K ctx
dense
Yi 1.5 6B6B10.7 GB84 tok/s4K ctx
dense
solar finalised finetuned Model 10.7B i1
C4810.7B13.9 GB46 tok/s280K ctx
dense
Yi 1.5 6B Chat
C486B10.4 GB84 tok/s565K ctx
dense
SOLAR 10.7B v1.0
C4810.7B13.9 GB46 tok/s280K ctx
dense
stablelm 2 1 6b chat imatrix
C486B10.4 GB84 tok/s565K ctx
dense
StarCoder2 7B7B10.8 GB96 tok/s16K ctx
dense
Helply 10.2b chat i1
C4810.2B13.5 GB49 tok/s298K ctx
dense
InternLM 20B20B38.8 GB26 tok/s8K ctx
dense
HelpingAI2 6B
C486B10.4 GB84 tok/s565K ctx
dense
Nemotron Mini 4B4B10.5 GB56 tok/s4K ctx
dense
HelpingAI2 6B i1
C486B10.4 GB84 tok/s565K ctx
dense
Qwen 2.5 Coder 0.5B0.5B6.6 GB7 tok/s131K ctx
dense
HelpingAI2.5 5B i1
C475B9.7 GB70 tok/s694K ctx
dense
Qwen3.5 4B4B9.0 GB56 tok/s889K ctx
dense
gemma 3 4b it
C464B9.0 GB56 tok/s889K ctx
dense
gemma 3 4b it
C464B9.0 GB56 tok/s889K ctx
dense
Qwen 3 0.6B0.6B7.3 GB8 tok/s33K ctx
dense
Llama 3.2 3B Instruct
C453B8.6 GB42 tok/s1198K ctx
dense
Llama 3.2 1B1B7.2 GB14 tok/s128K ctx
dense
Qwen2.5 3B Instruct3B8.3 GB42 tok/s1213K ctx
dense
Llama 3.2 3B Instruct
C453B8.3 GB42 tok/s1213K ctx
dense
Ministral 3 3B Instruct 2512
C453B8.3 GB42 tok/s1213K ctx
dense
HELVETE 3B
C443B8.3 GB42 tok/s1213K ctx
dense
Hermes 3 Llama 3.2 3B
C443B8.3 GB42 tok/s1213K ctx
dense
stablelm zephyr 3b
C443B8.3 GB42 tok/s1213K ctx
dense
StarCoder2 3B
C443B8.3 GB42 tok/s1213K ctx
dense
AI21 Jamba Reasoning 3B
C443B8.3 GB42 tok/s1213K ctx
dense
stablelm 3b 4e1t
C443B8.3 GB42 tok/s1213K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C443B8.3 GB42 tok/s1213K ctx
dense
ai21labs AI21 Jamba2 3B
C443B8.3 GB42 tok/s1213K ctx
dense
HelpingAI 3B hindi i1
C443B8.3 GB42 tok/s1213K ctx
dense
AI21 Jamba2 3B
C443B8.3 GB42 tok/s1213K ctx
dense
HelpingAI 3B hindi
C443B8.3 GB42 tok/s1213K ctx
dense
AI21 Jamba2 3B i1
C443B8.3 GB42 tok/s1213K ctx
dense
gemma 2 2b it
C432B8.0 GB28 tok/s1832K ctx
dense
EXAONE 3.5 2.4B Instruct
C432.4B7.8 GB34 tok/s1537K ctx
dense
StarCoder2 3B3B8.4 GB42 tok/s16K ctx
dense
gemma 2b2B7.5 GB28 tok/s1861K ctx
dense
Gemmasutra Mini 2B v1
C432B7.5 GB28 tok/s1861K ctx
dense
gemma 2 2b it
C432B7.5 GB28 tok/s1861K ctx
dense
Qwen 3.5 0.6B0.6B7.3 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B7.2 GB21 tok/s2509K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B7.2 GB21 tok/s2509K ctx
dense
Yi Coder 1.5B Chat
C421.5B7.2 GB21 tok/s2509K ctx
dense
stablelm 2 zephyr 1.6b
C411.6B7.2 GB22 tok/s2347K ctx
dense
logos16v2 stablelm2 1.6b i1
C411.6B7.2 GB22 tok/s2347K ctx
dense
Yi Coder 1.5B
C411.5B7.2 GB21 tok/s2509K ctx
dense
Falcon H1 1.5B Instruct1.5B7.2 GB21 tok/s2509K ctx
dense
TinyLlama 1.1B Chat v1.0
C411.1B6.9 GB15 tok/s3451K ctx
dense
EXAONE 4.0 1.2B
C411.2B7.0 GB17 tok/s3157K ctx
dense
Llama 3.2 1B Instruct Q8 0
C401B7.0 GB14 tok/s3776K ctx
dense
Qwen 2.5 0.5B0.5B6.6 GB7 tok/s131K ctx
dense
TinyLlama 1.1B Chat v0.3
C401.1B6.9 GB15 tok/s3451K ctx
dense
TinyLlama 1.1B Chat v0.61.1B6.9 GB15 tok/s3451K ctx
dense
Llama 3.2 1B Instruct
C401B6.8 GB14 tok/s3805K ctx
dense
gemma 3 1b it
C401B6.8 GB14 tok/s3805K ctx
dense
F
F Tier -- Can't Run (58 models)
Too heavy for this hardware.
TinyLlama 1.1B Chat v1.0 imatrix
D401.1B6.9 GB15 tok/s3451K ctx
dense
Falcon3 1B Instruct abliterated
D401B6.8 GB14 tok/s3805K ctx
dense
SmolVLM 500M Instruct
D380.5B6.6 GB7 tok/s4491K ctx
dense
embeddinggemma 300M
D370.3B6.4 GB4 tok/s4517K ctx
dense
granite embedding 107m multilingual
D340.11B6.2 GB2 tok/s4546K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B6.2 GB2 tok/s4547K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
D3248B41.0 GB16 tok/s4K ctx
dense
Llama 3.1 70B70B53.7 GB9 tok/s4K ctx
dense
Llama 3.3 70B70B53.7 GB9 tok/s4K ctx
dense
Qwen 2.5 72B72B54.9 GB8 tok/s4K ctx
dense
Qwen 3 235B A22B235B152.3 GB3 tok/s4K ctx
moe
Qwen3-Coder-Next80B56.3 GB12 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B301.8 GB3 tok/s4K ctx
moe
Qwen 3.5 122B A10B122B82.9 GB10 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B251.1 GB3 tok/s4K ctx
moe
Mixtral 8x22B141B95.5 GB4 tok/s4K ctx
moe
MPT-30B-Instruct30B51.1 GB17 tok/s5K ctx
dense
Devstral 2 123B Instruct123B86.5 GB4 tok/s4K ctx
dense
Mistral Small 4 119B119B84.0 GB11 tok/s4K ctx
moe
Leanstral 119B A6B119B87.5 GB10 tok/s4K ctx
moe
Mistral Large 3675B423.2 GB2 tok/s4K ctx
moe
GLM-5744B479.0 GB2 tok/s4K ctx
moe
GLM-5.1754B485.1 GB2 tok/s4K ctx
moe
Kimi K2.51000B623.5 GB2 tok/s4K ctx
moe
Kimi K2.61000B623.5 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B475.0 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B475.0 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B869.9 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B165.4 GB5 tok/s4K ctx
moe
DeepSeek R1 671B671B475.0 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B208.6 GB5 tok/s4K ctx
moe
DeepSeek V2.5 236B236B208.6 GB5 tok/s4K ctx
moe
DeepSeek LLM 67B67B52.8 GB9 tok/s4K ctx
dense
Llama 4 Scout 17B 16E109B75.5 GB6 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B253.0 GB3 tok/s4K ctx
moe
Command R+ 104B104B72.9 GB5 tok/s4K ctx
dense
Command A 111B111B77.7 GB4 tok/s4K ctx
dense
Pixtral Large 124B124B87.1 GB4 tok/s4K ctx
dense
Qwen 2.5 VL 72B72B54.9 GB8 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B54.9 GB8 tok/s4K ctx
dense
Nemotron 70B70B53.7 GB9 tok/s4K ctx
dense
Llama 3.1 405B405B260.8 GB2 tok/s4K ctx
dense
DeepSeek R1 Distill 70B70B53.7 GB9 tok/s4K ctx
dense
GPT-OSS 120B117B82.3 GB4 tok/s4K ctx
dense
DeepSeek V3.2671B415.9 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B150.2 GB6 tok/s4K ctx
moe
GLM-5.2753.3B484.6 GB2 tok/s4K ctx
moe
Solar Open 100B
F0100B78.8 GB4 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B177.7 GB2 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B57.0 GB7 tok/s4K ctx
dense
Baichuan M3 235B
F0235B177.0 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B177.0 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B58.4 GB7 tok/s4K ctx
dense
Solar Open 100B i1
F0100B78.8 GB4 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B56.3 GB8 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B57.0 GB7 tok/s4K ctx
dense
Qwen3.5 122B A10B122B80.2 GB4 tok/s4K ctx
dense
Qwen3.5 397B A17B397B294.8 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs