AI Compatibility Tier List
MacBook Pro M3 Pro 36GB
374 AI models ranked by how well they run on your 36 GB unified memory hardware. Best match: Qwen3-Coder 30B A3B Instruct (score: 92).
374
Models tested
304
Can run
36 GB
unified memory
17
Best tok/s
Tier Distribution
S 16
A 48
B 52
C 188
F 70
S: 16A: 48B: 52C: 188F: 70
S
S Tier -- Perfect Fit (16 models)
These models run flawlessly with headroom to spare.
Qwen3-Coder 30B A3B Instruct30.5B24.9 GB17 tok/s28K ctx
moe
Qwen3-VL 30B A3B Instruct30B24.6 GB17 tok/s31K ctx
moe
GPT-OSS 20B21B20.0 GB21 tok/s55K ctx
moe
Qwen 3 14B14B15.8 GB14 tok/s83K ctx
dense
Phi-4-reasoning-plus 14B14.7B16.8 GB13 tok/s33K ctx
dense
Qwen 3.5 9B9B12.5 GB21 tok/s114K ctx
dense
Qwen 3 30B A3B30.5B24.9 GB17 tok/s28K ctx
moe
Qwen 3.5 27B27B24.4 GB7 tok/s24K ctx
dense
Qwen 3.5 4B4B9.4 GB48 tok/s131K ctx
dense
Magistral Small 250724B21.9 GB8 tok/s43K ctx
dense
Qwen 3.6 27B27B22.2 GB6 tok/s76K ctx
dense
Devstral Small 2 24B Instruct24B21.9 GB8 tok/s43K ctx
dense
Qwen 3 8B8B11.9 GB24 tok/s118K ctx
dense
Nemotron Cascade 2 30B A3B30B26.0 GB17 tok/s15K ctx
moe
Devstral Small 1.124B21.9 GB8 tok/s43K ctx
dense
Nemotron 3 Nano 30B30B25.5 GB6 tok/s19K ctx
dense
A
A Tier -- Great Fit (48 models)
Excellent performance with comfortable memory margins.
Phi-4 Mini Reasoning 4B3.8B8.6 GB51 tok/s131K ctx
dense
Ministral 3 14B14B15.8 GB14 tok/s83K ctx
multimodal
Gemma 4 26B A4B25.2B23.8 GB18 tok/s25K ctx
moe
Codestral 2 25.0822B20.6 GB8 tok/s51K ctx
dense
CogVLM2 19B19B18.8 GB10 tok/s8K ctx
dense
Nemotron Nano 8B8B11.6 GB24 tok/s131K ctx
dense
DeepSeek Coder V2 16B16B17.8 GB27 tok/s55K ctx
moe
Phi-4 14B14B16.4 GB14 tok/s16K ctx
dense
Mistral Small 3.2 24B24B21.9 GB8 tok/s43K ctx
vision
Qwen 2.5 14B14B16.3 GB14 tok/s69K ctx
dense
Qwen 3.6 35B A3B35B30.2 GB12 tok/s4K ctx
moe
Qwen 3.5 35B A3B35B27.6 GB15 tok/s4K ctx
moe
InternVL2 8B8B11.6 GB24 tok/s8K ctx
dense
LFM2 24B24B21.9 GB8 tok/s43K ctx
dense
Granite Code 20B20B20.2 GB10 tok/s8K ctx
dense
Gemma 3 12B12B17.0 GB12 tok/s45K ctx
dense
Qwen 3 4B4B9.4 GB48 tok/s33K ctx
dense
Mistral Small 24B24B21.9 GB8 tok/s33K ctx
dense
Ministral 3 8B8B11.9 GB24 tok/s118K ctx
multimodal
MiniCPM-V 2.6 8B8B11.6 GB24 tok/s2K ctx
dense
Granite 4.1 30B30B27.0 GB6 tok/s12K ctx
dense
Jina Embeddings v30.57B7.9 GB8 tok/s8K ctx
dense
Mistral Small 3.1 24B24B21.9 GB8 tok/s43K ctx
dense
Nemotron Nano 9B v29B12.7 GB21 tok/s103K ctx
dense
SQLCoder 7B7B11.0 GB28 tok/s8K ctx
dense
BGE M30.57B7.1 GB8 tok/s8K ctx
dense
Qwen 3 32B32B28.2 GB5 tok/s7K ctx
dense
Qwen 2.5 VL 7B7B9.9 GB28 tok/s33K ctx
dense
OLMo 2 13B13B15.2 GB15 tok/s33K ctx
dense
CodeLlama 7B Instruct7B16.9 GB26 tok/s16K ctx
dense
Magistral 7B7B11.0 GB28 tok/s8K ctx
dense
StarCoder 7B7B16.4 GB26 tok/s8K ctx
dense
CodeGeeX 4 9B9B10.9 GB22 tok/s131K ctx
dense
DeepSeek R1 Distill 14B14B16.3 GB14 tok/s33K ctx
dense
Gemma 4 E4B8B10.9 GB18 tok/s128K ctx
dense
CodeLlama 13B Instruct13B24.9 GB14 tok/s16K ctx
dense
mxbai Embed Large0.34B6.9 GB5 tok/s1K ctx
dense
InternLM 7B7B16.9 GB26 tok/s8K ctx
dense
InternLM Chat 7B7B16.9 GB26 tok/s8K ctx
dense
Qwen 2.5 Coder 3B3B8.8 GB42 tok/s131K ctx
dense
Granite Code 8B8B11.6 GB24 tok/s8K ctx
dense
DevStral 7B7B11.0 GB28 tok/s8K ctx
dense
Qwen 2.5 7B7B9.9 GB28 tok/s131K ctx
dense
Pixtral 12B12B14.5 GB16 tok/s91K ctx
dense
Granite 4.1 8B8B12.1 GB24 tok/s107K ctx
dense
Snowflake Arctic Embed L0.34B6.9 GB5 tok/s1K ctx
dense
Codestral Mamba 7B7B9.5 GB30 tok/s262K ctx
state-space
LLaVA 1.6 13B13B24.9 GB14 tok/s4K ctx
dense
B
B Tier -- Good Fit (52 models)
Solid performance, may need quantization or reduced context.
EXAONE 4.0 32B32B28.2 GB5 tok/s7K ctx
dense
BGE Large EN v1.50.34B6.9 GB5 tok/s1K ctx
dense
LLaVA 1.5 7B7B16.9 GB26 tok/s4K ctx
dense
Nous Hermes 1.09B22.5 GB20 tok/s16K ctx
dense
Solar 7B7B12.0 GB28 tok/s8K ctx
dense
Qwen 2.5 32B32B28.2 GB5 tok/s7K ctx
dense
WizardLM 13B13B24.9 GB14 tok/s8K ctx
dense
Vicuna 13B13B24.9 GB14 tok/s4K ctx
dense
Llama 3.1 8B8B11.6 GB24 tok/s128K ctx
dense
Gemma 4 E2B5.1B8.4 GB29 tok/s128K ctx
dense
Ministral 3 3B3B7.4 GB42 tok/s262K ctx
multimodal
Nous Dolphin 13B13B26.4 GB12 tok/s15K ctx
dense
MPT-7B-Instruct7B16.9 GB26 tok/s8K ctx
dense
Baichuan 7B7B16.9 GB26 tok/s8K ctx
dense
Gemma 3 4B4B9.3 GB36 tok/s128K ctx
dense
WizardMath 7B7B11.0 GB28 tok/s4K ctx
dense
OLMo 2 7B7B11.0 GB28 tok/s4K ctx
dense
GLM-4 9B9B10.9 GB22 tok/s128K ctx
dense
Phi 4 Mini 4B4B8.7 GB48 tok/s128K ctx
dense
Phi 3 Mini 3.8B3.8B13.0 GB47 tok/s51K ctx
dense
Nomic Embed Text v1.50.14B5.6 GB2 tok/s8K ctx
dense
OLMo 2 32B32B28.2 GB5 tok/s4K ctx
dense
Qwen 3.5 2B2B7.7 GB28 tok/s131K ctx
dense
Qwen 2.5 Coder 7B7B9.9 GB28 tok/s131K ctx
dense
Phi 3.5 Mini 4B4B13.1 GB45 tok/s51K ctx
dense
Qwen 2.5 3B3B8.8 GB42 tok/s131K ctx
dense
DeepSeek R1 Distill 8B8B11.6 GB24 tok/s33K ctx
dense
Qwen 2.5 Coder 14B14B16.3 GB14 tok/s69K ctx
dense
Qwen 3 1.7B1.7B7.5 GB24 tok/s33K ctx
dense
Cerebras-GPT 13B13B23.9 GB12 tok/s19K ctx
dense
Gemma 2 9B9B15.4 GB16 tok/s8K ctx
dense
Baichuan 13B13B26.4 GB12 tok/s8K ctx
dense
Samantha 7B7B11.0 GB28 tok/s4K ctx
dense
DeepSeek R1 Distill 7B7B9.9 GB28 tok/s33K ctx
dense
Granite Code 3B3B9.1 GB42 tok/s8K ctx
dense
Granite 4.1 3B3B7.8 GB42 tok/s131K ctx
dense
Qwen 2.5 Coder 32B32B28.2 GB5 tok/s7K ctx
dense
Llama 3.2 11B Vision11B13.5 GB18 tok/s16K ctx
vision
Falcon 7B Instruct7B9.2 GB28 tok/s8K ctx
dense
All MiniLM L6 v20.02B5.1 GB2 tok/s0K ctx
dense
Granite Code 34B34B29.2 GB5 tok/s4K ctx
dense
StarCoder 15B15B30.2 GB8 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B6.1 GB21 tok/s33K ctx
dense
Command R 35B35B28.6 GB5 tok/s4K ctx
dense
Mistral Nemo 12B12B14.5 GB16 tok/s91K ctx
dense
Phi 3 Medium 14B14B16.4 GB14 tok/s66K ctx
dense
DeepSeek R1 Distill 32B32B28.2 GB5 tok/s7K ctx
dense
Codestral 22B22B20.6 GB9 tok/s33K ctx
dense
Mistral 7B Instruct v0.37B11.0 GB28 tok/s8K ctx
dense
Yi Coder 9B9B11.7 GB22 tok/s131K ctx
dense
Llama 3.2 3B3B8.3 GB42 tok/s128K ctx
dense
Ministral 8B8B11.9 GB24 tok/s118K ctx
dense
C
C Tier -- Tight Fit (188 models)
Runs with offloading or significant compromises.
SmolLM3 3B3B8.6 GB42 tok/s128K ctx
dense
DeepSeek R1 1.5B1.5B6.1 GB21 tok/s33K ctx
dense
TinyLlama 1.1B1.1B5.8 GB15 tok/s4K ctx
dense
Vicuna 7B7B16.9 GB26 tok/s4K ctx
dense
Granite 3.1 8B8B11.6 GB28 tok/s128K ctx
state-space
Yi 1.5 9B9B11.7 GB22 tok/s4K ctx
dense
Aya Expanse 32B32B26.7 GB6 tok/s8K ctx
dense
StarCoder2 15B15B16.8 GB11 tok/s16K ctx
dense
Qwen 2.5 Math 7B7B9.9 GB28 tok/s4K ctx
dense
Qwen 2.5 1.5B1.5B6.1 GB21 tok/s131K ctx
dense
DeepSeek LLM 7B7B16.4 GB26 tok/s4K ctx
dense
Codestral 22B v0.1
C4922B20.8 GB8 tok/s48K ctx
dense
Codestral 22B v0.1
C4922B20.8 GB8 tok/s48K ctx
dense
internlm2 5 20b chat
C4920B19.3 GB9 tok/s61K ctx
dense
OpenChat 7B7B11.0 GB28 tok/s8K ctx
dense
Codestral RAG 19B Pruned i1
C4919B18.6 GB9 tok/s69K ctx
dense
internlm2 limarp chat 20b
C4920B19.3 GB9 tok/s61K ctx
dense
Gemma 2 2B2B7.6 GB28 tok/s8K ctx
dense
internlm2 math plus 20b i1
C4920B19.3 GB9 tok/s61K ctx
dense
Codestral 22B v0.1
C4922B20.8 GB8 tok/s48K ctx
dense
Codestral 21B Pruned i1
C4921B20.1 GB9 tok/s54K ctx
dense
Codestral 22B v0.1 IMat
C4922B20.8 GB8 tok/s48K ctx
dense
Aya Expanse 8B8B11.6 GB24 tok/s8K ctx
dense
Codestral 22B v0.1 i1
C4922B20.8 GB8 tok/s48K ctx
dense
Gemma 3 1B1B5.8 GB14 tok/s33K ctx
dense
Phi 4 reasoning vision 15B
C4815B15.7 GB12 tok/s109K ctx
dense
StarCoder2 15B
C4815B15.7 GB12 tok/s109K ctx
dense
Nemotron Mini 4B4B9.2 GB48 tok/s4K ctx
dense
DeepSeek R1 Distill Qwen 14B14B15.0 GB13 tok/s123K ctx
dense
starcoder2 15b instruct v0.1
C4815B15.7 GB12 tok/s109K ctx
dense
starcoder2 15b i1
C4815B15.7 GB12 tok/s109K ctx
dense
starcoder2 15b instruct v0.1
C4815B15.7 GB12 tok/s109K ctx
dense
HelpingAI 15B i1
C4815B15.7 GB12 tok/s109K ctx
dense
OpenHermes 2.5 7B7B11.0 GB28 tok/s8K ctx
dense
NousResearch Hermes 4 14B
C4814B15.0 GB13 tok/s123K ctx
dense
Dolphin 2.9 8B8B11.6 GB24 tok/s33K ctx
dense
Zephyr 7B Beta7B11.0 GB28 tok/s33K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4814B15.0 GB13 tok/s123K ctx
dense
internlm JanusCoder 14B
C4814B15.0 GB13 tok/s123K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4814B15.0 GB13 tok/s123K ctx
dense
Starling LM 7B7B11.0 GB28 tok/s8K ctx
dense
gemma 3 12b it
C4712B13.5 GB15 tok/s157K ctx
dense
Yi 1.5 34B34B29.2 GB5 tok/s4K ctx
dense
Qwen 2.5 Coder 0.5B0.5B5.3 GB7 tok/s131K ctx
dense
StableLM 2 12B12B25.6 GB12 tok/s4K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4710.7B12.6 GB17 tok/s186K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4710.7B12.6 GB17 tok/s186K ctx
dense
Qwen3.5 9B9B11.3 GB20 tok/s237K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C479B11.3 GB20 tok/s237K ctx
dense
Qwen3.5 9B
C479B11.3 GB20 tok/s237K ctx
dense
solar finalised finetuned Model 10.7B i1
C4610.7B12.6 GB17 tok/s186K ctx
dense
SOLAR 10.7B v1.0
C4610.7B12.6 GB17 tok/s186K ctx
dense
Meta Llama 3.1 8B Instruct
C468B10.6 GB22 tok/s277K ctx
dense
llava llama 3 8b v1 1
C468B10.6 GB22 tok/s277K ctx
dense
Yi Coder 9B Chat
C469B11.3 GB20 tok/s237K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B10.6 GB22 tok/s277K ctx
dense
Neural Chat 7B7B11.0 GB28 tok/s8K ctx
dense
glm 4 9b chat 1m
C469B11.3 GB20 tok/s237K ctx
dense
Meta Llama 3 8B Instruct
C468B10.6 GB22 tok/s277K ctx
dense
DeepSeek R1 Distill Llama 8B8B10.6 GB22 tok/s277K ctx
dense
Dolphin3.0 Llama3.1 8B
C468B10.6 GB22 tok/s277K ctx
dense
Llama 3 8B Instruct 32k v0.1
C468B10.6 GB22 tok/s277K ctx
dense
Llama 2 7B Chat
C467B9.9 GB26 tok/s329K ctx
dense
Meta Llama 3.1 8B Instruct
C468B10.6 GB22 tok/s277K ctx
dense
vntl llama3 8b v2
C468B10.6 GB22 tok/s277K ctx
dense
falcon mamba 7b instruct Q4 K M
C467B9.9 GB30 tok/s329K ctx
dense
Qwen3.5 4B4B7.7 GB45 tok/s638K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C468B10.6 GB22 tok/s277K ctx
dense
Mistral 7B Instruct v0.2
C467B9.9 GB26 tok/s329K ctx
dense
Yi 1.5 9B Chat
C469B11.3 GB20 tok/s237K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C468B10.6 GB22 tok/s277K ctx
dense
Mistral 7B Instruct v0.3
C467B9.9 GB26 tok/s329K ctx
dense
gemma 3 4b it
C464B7.7 GB45 tok/s638K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C468B10.6 GB22 tok/s277K ctx
dense
HelpingAI2.5 10B i1
C4610B12.1 GB18 tok/s205K ctx
dense
Helply 10.2b chat i1
C4610.2B12.2 GB18 tok/s200K ctx
dense
Hermes 3 Llama 3.1 8B
C468B10.6 GB22 tok/s277K ctx
dense
Mistral Small 3.2 24B Instruct 250624B22.2 GB8 tok/s37K ctx
dense
Yi 1.5 6B6B9.4 GB33 tok/s4K ctx
dense
gemma 3 4b it
C464B7.7 GB45 tok/s638K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B22.2 GB8 tok/s37K ctx
dense
Mistral Small 24B Instruct 2501
C4624B22.2 GB8 tok/s37K ctx
dense
Hermes 2 Pro Llama 3 8B
C468B10.6 GB22 tok/s277K ctx
dense
Dolphin3.0 Llama3.1 8B
C468B10.6 GB22 tok/s277K ctx
dense
Mamba Codestral 7B v0.1
C467B9.9 GB30 tok/s329K ctx
dense
granite 8b code instruct 4k
C468B10.6 GB22 tok/s277K ctx
dense
dolphin 2.9.4 llama3.1 8b
C468B10.6 GB22 tok/s277K ctx
dense
Qwen3.5 27B27B24.4 GB7 tok/s24K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4624B22.2 GB8 tok/s37K ctx
dense
Yi 9B Coder i1
C469B11.3 GB20 tok/s237K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4624B22.2 GB8 tok/s37K ctx
dense
Mistral 7B Instruct v0.3
C467B9.9 GB26 tok/s329K ctx
dense
Hermes 2 Pro Mistral 7B
C467B9.9 GB26 tok/s329K ctx
dense
Yi 1.5 6B Chat
C466B9.2 GB30 tok/s398K ctx
dense
openchat 3.6 8b 20240522 IMat
C468B10.6 GB22 tok/s277K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B10.5 GB23 tok/s287K ctx
dense
zephyr 7B beta
C467B9.9 GB26 tok/s329K ctx
dense
aya expanse 8b
C468B10.6 GB22 tok/s277K ctx
dense
HelpingAI2 9B
C469B11.3 GB20 tok/s237K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C467B9.9 GB26 tok/s329K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4624B22.2 GB8 tok/s37K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B10.5 GB23 tok/s287K ctx
dense
stablelm 2 zephyr 1 6b
C466B9.2 GB30 tok/s398K ctx
dense
dolphin v2 8b abliterated i1
C468B10.6 GB22 tok/s277K ctx
dense
gemma 3 27b it27B24.4 GB7 tok/s24K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C467.8B10.5 GB23 tok/s287K ctx
dense
Falcon H1R 7B
C467B9.9 GB26 tok/s329K ctx
dense
HelpingAI2 9B i1
C469B11.3 GB20 tok/s237K ctx
dense
HelpingAI 9B 200k i1
C469B11.3 GB20 tok/s237K ctx
dense
exaone 3.0 7.8b it
C467.8B10.5 GB23 tok/s287K ctx
dense
aya expanse 8b orthogonal heretic i1
C468B10.6 GB22 tok/s277K ctx
dense
zephyr 7B alpha
C467B9.9 GB26 tok/s329K ctx
dense
speechless zephyr code functionary 7b
C467B9.9 GB26 tok/s329K ctx
dense
HelpingAI 9B i1
C469B11.3 GB20 tok/s237K ctx
dense
StarCoder2 7B
C467B9.9 GB26 tok/s329K ctx
dense
Llama 3.2 3B Instruct
C463B7.3 GB42 tok/s863K ctx
dense
japanese stablelm instruct gamma 7B
C467B9.9 GB26 tok/s329K ctx
dense
aya expanse 8b orthogonal heretic
C468B10.6 GB22 tok/s277K ctx
dense
Yi 1.5 6B Chat
C466B9.2 GB30 tok/s398K ctx
dense
Falcon H1 7B Instruct
C467B9.9 GB26 tok/s329K ctx
dense
Falcon H1R 7B7B9.9 GB26 tok/s329K ctx
dense
stablelm 2 1 6b chat imatrix
C466B9.2 GB30 tok/s398K ctx
dense
gemma 3 27b it
C4627B24.4 GB7 tok/s24K ctx
dense
internlm2 5 1 8b chat i1
C468B10.6 GB22 tok/s277K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B22.2 GB8 tok/s37K ctx
dense
internlm3 8b instruct abliterated i1
C468B10.6 GB22 tok/s277K ctx
dense
stabilityai japanese stablelm base gamma 7b
C467B9.9 GB26 tok/s329K ctx
dense
baichuan2 7b chat
C467B9.9 GB26 tok/s329K ctx
dense
DiscoPOP zephyr 7b gemma
C467B9.9 GB26 tok/s329K ctx
dense
starcoder2 7b
C467B9.9 GB26 tok/s329K ctx
dense
Qwen 3 0.6B0.6B6.0 GB8 tok/s33K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C457B9.9 GB26 tok/s329K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C457B9.9 GB26 tok/s329K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C457B9.9 GB26 tok/s329K ctx
dense
internlm2 math plus 7b IMat
C457B9.9 GB26 tok/s329K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C457B9.9 GB26 tok/s329K ctx
dense
internlm2 5 7b chat i1
C457B9.9 GB26 tok/s329K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C457B9.9 GB26 tok/s329K ctx
dense
MD Judge v0 2 internlm2 7b i1
C457B9.9 GB26 tok/s329K ctx
dense
Qwen2.5 3B Instruct3B7.0 GB42 tok/s878K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C457B9.9 GB26 tok/s329K ctx
dense
jointpreferences mistral 7b sft helpful
C457B9.9 GB26 tok/s329K ctx
dense
zephyr 7b dpo full i1
C457B9.9 GB26 tok/s329K ctx
dense
HelpingAI2 6B
C456B9.2 GB30 tok/s398K ctx
dense
blossom v3 baichuan2 7b i1
C457B9.9 GB26 tok/s329K ctx
dense
blossom v1 baichuan 7b i1
C457B9.9 GB26 tok/s329K ctx
dense
HelpingAI2 6B i1
C456B9.2 GB30 tok/s398K ctx
dense
Llama 3.2 1B1B5.9 GB14 tok/s128K ctx
dense
HelpingAI2.5 5B i1
C455B8.4 GB36 tok/s494K ctx
dense
Llama 3.2 3B Instruct
C453B7.0 GB42 tok/s878K ctx
dense
Ministral 3 3B Instruct 2512
C453B7.0 GB42 tok/s878K ctx
dense
StarCoder2 7B7B9.5 GB28 tok/s16K ctx
dense
HELVETE 3B
C453B7.0 GB42 tok/s878K ctx
dense
Hermes 3 Llama 3.2 3B
C453B7.0 GB42 tok/s878K ctx
dense
stablelm zephyr 3b
C453B7.0 GB42 tok/s878K ctx
dense
StarCoder2 3B
C453B7.0 GB42 tok/s878K ctx
dense
AI21 Jamba Reasoning 3B
C453B7.0 GB42 tok/s878K ctx
dense
stablelm 3b 4e1t
C453B7.0 GB42 tok/s878K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C453B7.0 GB42 tok/s878K ctx
dense
ai21labs AI21 Jamba2 3B
C453B7.0 GB42 tok/s878K ctx
dense
HelpingAI 3B hindi i1
C453B7.0 GB42 tok/s878K ctx
dense
AI21 Jamba2 3B
C453B7.0 GB42 tok/s878K ctx
dense
HelpingAI 3B hindi
C453B7.0 GB42 tok/s878K ctx
dense
AI21 Jamba2 3B i1
C453B7.0 GB42 tok/s878K ctx
dense
gemma 2 2b it
C442B6.7 GB28 tok/s1331K ctx
dense
EXAONE 3.5 2.4B Instruct
C442.4B6.5 GB34 tok/s1119K ctx
dense
StarCoder2 3B3B7.1 GB42 tok/s16K ctx
dense
gemma 2b2B6.2 GB28 tok/s1359K ctx
dense
Gemmasutra Mini 2B v1
C432B6.2 GB28 tok/s1359K ctx
dense
gemma 2 2b it
C432B6.2 GB28 tok/s1359K ctx
dense
Qwen 3.5 0.6B0.6B6.0 GB8 tok/s131K ctx
dense
Qwen2.5 1.5B Instruct1.5B5.9 GB21 tok/s1840K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B5.9 GB21 tok/s1840K ctx
dense
Yi Coder 1.5B Chat
C421.5B5.9 GB21 tok/s1840K ctx
dense
stablelm 2 zephyr 1.6b
C421.6B6.0 GB22 tok/s1720K ctx
dense
logos16v2 stablelm2 1.6b i1
C421.6B6.0 GB22 tok/s1720K ctx
dense
Yi Coder 1.5B
C421.5B5.9 GB21 tok/s1840K ctx
dense
Falcon H1 1.5B Instruct1.5B5.9 GB21 tok/s1840K ctx
dense
TinyLlama 1.1B Chat v1.0
C411.1B5.6 GB15 tok/s2540K ctx
dense
EXAONE 4.0 1.2B
C411.2B5.7 GB17 tok/s2321K ctx
dense
Llama 3.2 1B Instruct Q8 0
C411B5.7 GB14 tok/s2773K ctx
dense
Qwen 2.5 0.5B0.5B5.3 GB7 tok/s131K ctx
dense
TinyLlama 1.1B Chat v0.3
C411.1B5.6 GB15 tok/s2540K ctx
dense
TinyLlama 1.1B Chat v0.61.1B5.6 GB15 tok/s2540K ctx
dense
Llama 3.2 1B Instruct
C411B5.5 GB14 tok/s2802K ctx
dense
gemma 3 1b it
C401B5.5 GB14 tok/s2802K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
C401.1B5.6 GB15 tok/s2540K ctx
dense
Falcon3 1B Instruct abliterated
C401B5.5 GB14 tok/s2802K ctx
dense
F
F Tier -- Can't Run (70 models)
Too heavy for this hardware.
SmolVLM 500M Instruct
D380.5B5.3 GB7 tok/s3316K ctx
dense
embeddinggemma 300M
D370.3B5.1 GB4 tok/s3342K ctx
dense
Yi 34B Chat34B29.2 GB5 tok/s4K ctx
dense
Qwen3.5 35B A3B35B30.2 GB4 tok/s4K ctx
dense
EXAONE 4.0 32B
D3532B28.1 GB5 tok/s7K ctx
dense
granite embedding 107m multilingual
D350.11B5.0 GB2 tok/s3371K ctx
dense
aya expanse 32b heretic MPOA i1
D3532B28.1 GB5 tok/s7K ctx
dense
Baichuan M2 32B Q4 K M
D3432B28.1 GB5 tok/s7K ctx
dense
Qwen3.5 35B A3B
D3435B30.2 GB4 tok/s4K ctx
dense
baichuan inc Baichuan M2 32B
D3432B28.1 GB5 tok/s7K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B4.9 GB2 tok/s3372K ctx
dense
Hermes 4.3 36B
D2836B31.0 GB4 tok/s4K ctx
dense
Llama 3.1 70B70B52.4 GB2 tok/s4K ctx
dense
Llama 3.3 70B70B52.4 GB2 tok/s4K ctx
dense
Qwen 2.5 72B72B53.6 GB2 tok/s4K ctx
dense
Qwen 3 235B A22B235B151.0 GB2 tok/s4K ctx
moe
Qwen3-Coder-Next80B55.1 GB4 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B300.5 GB2 tok/s4K ctx
moe
Qwen 3.5 122B A10B122B81.6 GB3 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B249.8 GB2 tok/s4K ctx
moe
Gemma 2 27B27B32.5 GB4 tok/s7K ctx
dense
Gemma 3 27B27B32.5 GB4 tok/s7K ctx
dense
Mixtral 8x7B47B35.4 GB5 tok/s4K ctx
moe
Mixtral 8x22B141B94.2 GB2 tok/s4K ctx
moe
Falcon 40B Instruct40B35.4 GB3 tok/s4K ctx
dense
MPT-30B-Instruct30B49.8 GB2 tok/s4K ctx
dense
Devstral 2 123B Instruct123B85.2 GB2 tok/s4K ctx
dense
Mistral Small 4 119B119B82.7 GB3 tok/s4K ctx
moe
Leanstral 119B A6B119B86.2 GB3 tok/s4K ctx
moe
Mistral Large 3675B421.9 GB2 tok/s4K ctx
moe
GLM-5744B477.7 GB2 tok/s4K ctx
moe
GLM-5.1754B483.8 GB2 tok/s4K ctx
moe
Kimi K2.51000B622.2 GB2 tok/s4K ctx
moe
Kimi K2.61000B622.2 GB2 tok/s4K ctx
moe
Kimi Linear 48B A3B48B35.0 GB3 tok/s4K ctx
linear
InternLM 20B20B37.5 GB5 tok/s6K ctx
dense
DeepSeek V3 671B671B473.7 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B473.7 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B868.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B164.1 GB2 tok/s4K ctx
moe
DeepSeek R1 671B671B473.7 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B207.3 GB2 tok/s4K ctx
moe
DeepSeek V2.5 236B236B207.3 GB2 tok/s4K ctx
moe
DeepSeek LLM 67B67B51.5 GB2 tok/s4K ctx
dense
Llama 4 Scout 17B 16E109B74.2 GB2 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B251.7 GB2 tok/s4K ctx
moe
Command R+ 104B104B71.6 GB2 tok/s4K ctx
dense
Command A 111B111B76.4 GB2 tok/s4K ctx
dense
Pixtral Large 124B124B85.8 GB2 tok/s4K ctx
dense
Qwen 2.5 VL 72B72B53.6 GB2 tok/s4K ctx
dense
Qwen 2.5 Math 72B72B53.6 GB2 tok/s4K ctx
dense
Nemotron 70B70B52.4 GB2 tok/s4K ctx
dense
Llama 3.1 405B405B259.5 GB2 tok/s4K ctx
dense
DeepSeek R1 Distill 70B70B52.4 GB2 tok/s4K ctx
dense
GPT-OSS 120B117B81.0 GB2 tok/s4K ctx
dense
Gemma 4 31B30.7B38.2 GB3 tok/s4K ctx
dense
DeepSeek V3.2671B414.6 GB2 tok/s4K ctx
moe
MiniMax M2.7
F0230B148.9 GB2 tok/s4K ctx
moe
Solar Open 100B
F0100B77.5 GB2 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
F048B39.7 GB2 tok/s4K ctx
dense
K EXAONE 236B A23B
F0236B176.4 GB2 tok/s4K ctx
dense
Llama 3.3 70B Instruct
F070B55.7 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B175.7 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B175.7 GB2 tok/s4K ctx
dense
BaichuanMed OCR 72B i1
F072B57.1 GB2 tok/s4K ctx
dense
Solar Open 100B i1
F0100B77.5 GB2 tok/s4K ctx
dense
Solar Open 69B REAP i1
F069B55.0 GB2 tok/s4K ctx
dense
stabilityai japanese stablelm instruct beta 70b
F070B55.7 GB2 tok/s4K ctx
dense
Qwen3.5 122B A10B122B78.9 GB2 tok/s4K ctx
dense
Qwen3.5 397B A17B397B293.5 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs