AI Compatibility Tier List
Mac Studio M2 Ultra 128GB
374 AI models ranked by how well they run on your 128 GB unified memory hardware. Best match: Qwen3-Coder-Next (score: 93).
374
Models tested
338
Can run
128 GB
unified memory
31
Best tok/s
Tier Distribution
S 26
A 61
B 56
C 195
F 36
S: 26A: 61B: 56C: 195F: 36
S
S Tier -- Perfect Fit (26 models)
These models run flawlessly with headroom to spare.
Qwen3-Coder-Next80B65.0 GB31 tok/s256K ctx
moe
Qwen 3.5 122B A10B122B91.6 GB29 tok/s20K ctx
moe
Qwen3-Coder 30B A3B Instruct30.5B34.8 GB70 tok/s256K ctx
moe
Qwen 3.6 35B A3B35B40.2 GB59 tok/s219K ctx
moe
Qwen3-VL 30B A3B Instruct30B34.5 GB73 tok/s256K ctx
moe
Qwen 3.5 35B A3B35B37.5 GB64 tok/s131K ctx
moe
Mistral Small 4 119B119B92.7 GB31 tok/s14K ctx
moe
Qwen 2.5 VL 72B72B63.5 GB12 tok/s33K ctx
dense
Qwen 3 30B A3B30.5B34.8 GB70 tok/s131K ctx
moe
Qwen 3.5 27B27B34.4 GB30 tok/s131K ctx
dense
Qwen 3.5 9B9B22.4 GB91 tok/s131K ctx
dense
Devstral 2 123B Instruct123B95.1 GB6 tok/s7K ctx
dense
Qwen 3 32B32B38.2 GB26 tok/s131K ctx
dense
Magistral Small 250724B31.8 GB34 tok/s131K ctx
dense
Devstral Small 2 24B Instruct24B31.8 GB34 tok/s256K ctx
dense
Qwen 3.6 27B27B32.2 GB23 tok/s262K ctx
dense
GPT-OSS 20B21B30.0 GB89 tok/s128K ctx
moe
Nemotron Cascade 2 30B A3B30B36.0 GB72 tok/s262K ctx
moe
Qwen 3 14B14B25.7 GB59 tok/s131K ctx
dense
Qwen 3 8B8B21.8 GB102 tok/s131K ctx
dense
Command A 111B111B86.3 GB8 tok/s40K ctx
dense
Nemotron 3 Nano 30B30B35.5 GB27 tok/s131K ctx
dense
GPT-OSS 120B117B91.0 GB7 tok/s20K ctx
dense
Phi-4-reasoning-plus 14B14.7B26.7 GB56 tok/s33K ctx
dense
Leanstral 119B A6B119B96.1 GB27 tok/s9K ctx
moe
Devstral Small 1.124B31.8 GB34 tok/s131K ctx
dense
A
A Tier -- Great Fit (61 models)
Excellent performance with comfortable memory margins.
Pixtral Large 124B124B95.7 GB6 tok/s5K ctx
dense
Gemma 4 31B30.7B48.1 GB20 tok/s64K ctx
dense
Qwen 3.5 4B4B19.4 GB56 tok/s131K ctx
dense
Gemma 4 26B A4B25.2B33.8 GB75 tok/s256K ctx
moe
Llama 3.3 70B70B62.3 GB12 tok/s114K ctx
dense
Nemotron Nano 8B8B21.6 GB102 tok/s131K ctx
dense
EXAONE 4.0 32B32B38.2 GB26 tok/s131K ctx
dense
Phi-4 Mini Reasoning 4B3.8B18.5 GB53 tok/s131K ctx
dense
Ministral 3 14B14B25.7 GB58 tok/s262K ctx
multimodal
Qwen 2.5 72B72B63.5 GB12 tok/s110K ctx
dense
Llama 3.1 70B70B62.3 GB12 tok/s114K ctx
dense
Qwen 2.5 32B32B38.2 GB26 tok/s131K ctx
dense
Codestral 2 25.0822B30.6 GB35 tok/s256K ctx
dense
Mistral Small 3.2 24B24B31.8 GB34 tok/s131K ctx
vision
Gemma 3 27B27B42.4 GB22 tok/s87K ctx
dense
CogVLM2 19B19B28.8 GB43 tok/s8K ctx
dense
LFM2 24B24B31.8 GB34 tok/s131K ctx
dense
InternVL2 8B8B21.6 GB102 tok/s8K ctx
dense
Kimi Linear 48B A3B48B44.9 GB16 tok/s831K ctx
linear
OLMo 2 32B32B38.2 GB26 tok/s4K ctx
dense
Granite 4.1 30B30B36.9 GB27 tok/s131K ctx
dense
Phi-4 14B14B26.3 GB58 tok/s16K ctx
dense
Mistral Small 24B24B31.8 GB34 tok/s33K ctx
dense
Ministral 3 8B8B21.8 GB102 tok/s262K ctx
multimodal
MiniCPM-V 2.6 8B8B21.6 GB102 tok/s2K ctx
dense
Qwen 2.5 14B14B26.2 GB59 tok/s131K ctx
dense
Mistral Small 3.1 24B24B31.8 GB34 tok/s131K ctx
dense
DeepSeek Coder V2 16B16B27.8 GB113 tok/s131K ctx
moe
Nemotron Nano 9B v29B22.7 GB91 tok/s131K ctx
dense
Llama 4 Scout 17B 16E109B84.1 GB18 tok/s60K ctx
moe
SQLCoder 7B7B20.9 GB98 tok/s8K ctx
dense
Qwen 3 4B4B19.4 GB56 tok/s33K ctx
dense
DeepSeek R1 Distill 70B70B62.3 GB12 tok/s114K ctx
dense
Qwen 2.5 VL 7B7B19.8 GB98 tok/s33K ctx
dense
Granite Code 20B20B30.1 GB41 tok/s8K ctx
dense
Gemma 3 12B12B26.9 GB50 tok/s131K ctx
dense
Jina Embeddings v30.57B17.8 GB8 tok/s8K ctx
dense
Magistral 7B7B20.9 GB98 tok/s8K ctx
dense
CodeGeeX 4 9B9B20.8 GB92 tok/s131K ctx
dense
Qwen 2.5 Coder 32B32B38.2 GB26 tok/s131K ctx
dense
BGE M30.57B17.0 GB8 tok/s8K ctx
dense
StarCoder 15B15B40.2 GB44 tok/s8K ctx
dense
CodeLlama 13B Instruct13B34.9 GB59 tok/s16K ctx
dense
Granite Code 34B34B39.1 GB24 tok/s8K ctx
dense
Gemma 4 E4B8B20.9 GB78 tok/s128K ctx
dense
OLMo 2 13B13B25.1 GB63 tok/s33K ctx
dense
Command R 35B35B38.5 GB24 tok/s131K ctx
dense
Qwen 2.5 7B7B19.8 GB98 tok/s131K ctx
dense
Granite Code 8B8B21.6 GB102 tok/s8K ctx
dense
DeepSeek R1 Distill 32B32B38.2 GB26 tok/s33K ctx
dense
LLaVA 1.6 13B13B34.9 GB59 tok/s4K ctx
dense
DevStral 7B7B20.9 GB98 tok/s8K ctx
dense
MPT-30B-Instruct30B59.8 GB22 tok/s8K ctx
dense
Codestral Mamba 7B7B19.5 GB98 tok/s262K ctx
state-space
Nemotron 70B70B62.3 GB12 tok/s114K ctx
dense
CodeLlama 7B Instruct7B26.8 GB98 tok/s16K ctx
dense
StarCoder 7B7B26.3 GB98 tok/s8K ctx
dense
Granite 4.1 8B8B22.0 GB102 tok/s131K ctx
dense
DeepSeek R1 Distill 14B14B26.2 GB59 tok/s33K ctx
dense
mxbai Embed Large0.34B16.9 GB5 tok/s1K ctx
dense
Pixtral 12B12B24.5 GB68 tok/s131K ctx
dense
B
B Tier -- Good Fit (56 models)
Solid performance, may need quantization or reduced context.
Snowflake Arctic Embed L0.34B16.9 GB5 tok/s1K ctx
dense
Nous Hermes 1.09B32.4 GB85 tok/s16K ctx
dense
Qwen 2.5 Coder 3B3B18.8 GB42 tok/s131K ctx
dense
WizardLM 13B13B34.9 GB59 tok/s8K ctx
dense
Vicuna 13B13B34.9 GB59 tok/s4K ctx
dense
Nous Dolphin 13B13B36.3 GB51 tok/s16K ctx
dense
InternLM 7B7B26.8 GB98 tok/s8K ctx
dense
InternLM Chat 7B7B26.8 GB98 tok/s8K ctx
dense
Gemma 4 E2B5.1B18.4 GB71 tok/s128K ctx
dense
GLM-4 9B9B20.8 GB92 tok/s128K ctx
dense
BGE Large EN v1.50.34B16.9 GB5 tok/s1K ctx
dense
Llama 3.1 8B8B21.6 GB102 tok/s128K ctx
dense
Falcon 40B Instruct40B45.4 GB18 tok/s8K ctx
dense
Solar 7B7B21.9 GB98 tok/s8K ctx
dense
WizardMath 7B7B20.9 GB98 tok/s4K ctx
dense
OLMo 2 7B7B20.9 GB98 tok/s4K ctx
dense
Gemma 2 27B27B42.4 GB22 tok/s8K ctx
dense
Nomic Embed Text v1.50.14B15.6 GB2 tok/s8K ctx
dense
Qwen 2.5 Coder 7B7B19.8 GB98 tok/s131K ctx
dense
LLaVA 1.5 7B7B26.8 GB98 tok/s4K ctx
dense
Gemma 3 4B4B19.2 GB56 tok/s128K ctx
dense
Ministral 3 3B3B17.3 GB42 tok/s262K ctx
multimodal
Phi 4 Mini 4B4B18.6 GB56 tok/s128K ctx
dense
Mixtral 8x7B47B45.3 GB33 tok/s33K ctx
moe
Baichuan 13B13B36.3 GB51 tok/s8K ctx
dense
Qwen 3.5 2B2B17.7 GB28 tok/s131K ctx
dense
MPT-7B-Instruct7B26.8 GB98 tok/s8K ctx
dense
Baichuan 7B7B26.8 GB98 tok/s8K ctx
dense
Cerebras-GPT 13B13B33.8 GB51 tok/s112K ctx
dense
DeepSeek R1 Distill 7B7B19.8 GB98 tok/s33K ctx
dense
DeepSeek R1 Distill 8B8B21.6 GB102 tok/s33K ctx
dense
Command R+ 104B104B81.6 GB8 tok/s66K ctx
dense
Qwen 2.5 Math 72B72B63.5 GB12 tok/s4K ctx
dense
Samantha 7B7B20.9 GB98 tok/s4K ctx
dense
Qwen 2.5 3B3B18.8 GB42 tok/s131K ctx
dense
Phi 3 Mini 3.8B3.8B22.9 GB53 tok/s128K ctx
dense
Falcon 7B Instruct7B19.1 GB98 tok/s8K ctx
dense
Qwen 3 1.7B1.7B17.5 GB24 tok/s33K ctx
dense
Phi 3.5 Mini 4B4B23.0 GB56 tok/s128K ctx
dense
Gemma 2 9B9B25.3 GB67 tok/s8K ctx
dense
Llama 3.2 11B Vision11B23.4 GB74 tok/s16K ctx
vision
All MiniLM L6 v20.02B15.0 GB2 tok/s0K ctx
dense
Qwen 2.5 Coder 14B14B26.2 GB59 tok/s131K ctx
dense
Granite 4.1 3B3B17.8 GB42 tok/s131K ctx
dense
Granite Code 3B3B19.0 GB42 tok/s8K ctx
dense
Qwen 2.5 Coder 1.5B1.5B16.1 GB21 tok/s33K ctx
dense
DeepSeek LLM 67B67B61.4 GB12 tok/s4K ctx
dense
Yi 1.5 34B34B39.1 GB24 tok/s4K ctx
dense
Yi Coder 9B9B21.7 GB92 tok/s131K ctx
dense
Mistral Nemo 12B12B24.5 GB68 tok/s128K ctx
dense
Mistral 7B Instruct v0.37B20.9 GB98 tok/s8K ctx
dense
Phi 3 Medium 14B14B26.3 GB58 tok/s128K ctx
dense
InternLM 20B20B47.4 GB33 tok/s8K ctx
dense
Ministral 8B8B21.8 GB102 tok/s131K ctx
dense
Codestral 22B22B30.6 GB37 tok/s33K ctx
dense
Llama 3.2 3B3B18.3 GB42 tok/s128K ctx
dense
C
C Tier -- Tight Fit (195 models)
Runs with offloading or significant compromises.
Aya Expanse 32B32B36.7 GB26 tok/s8K ctx
dense
DeepSeek R1 1.5B1.5B16.1 GB21 tok/s33K ctx
dense
Granite 3.1 8B8B21.6 GB112 tok/s128K ctx
state-space
TinyLlama 1.1B1.1B15.7 GB15 tok/s4K ctx
dense
SmolLM3 3B3B18.5 GB42 tok/s128K ctx
dense
Yi 1.5 9B9B21.7 GB92 tok/s4K ctx
dense
Qwen 2.5 Math 7B7B19.8 GB98 tok/s4K ctx
dense
Llama 3.3 70B Instruct
C5070B65.6 GB11 tok/s68K ctx
dense
Mixtral 8x22B141B104.2 GB9 tok/s4K ctx
moe
Solar Open 69B REAP i1
C5069B64.9 GB11 tok/s70K ctx
dense
stabilityai japanese stablelm instruct beta 70b
C5070B65.6 GB11 tok/s68K ctx
dense
BaichuanMed OCR 72B i1
C4972B67.1 GB11 tok/s64K ctx
dense
Qwen 2.5 1.5B1.5B16.1 GB21 tok/s131K ctx
dense
OpenChat 7B7B20.9 GB98 tok/s8K ctx
dense
Vicuna 7B7B26.8 GB98 tok/s4K ctx
dense
Aya Expanse 8B8B21.6 GB102 tok/s8K ctx
dense
Yi 34B Chat34B39.1 GB24 tok/s200K ctx
dense
StableLM 2 12B12B35.6 GB50 tok/s4K ctx
dense
Qwen3 48B A4B Savant Commander Distill 12X Closed Open Heretic Uncensored
C4848B49.6 GB16 tok/s137K ctx
dense
Dolphin 2.9 8B8B21.6 GB102 tok/s33K ctx
dense
OpenHermes 2.5 7B7B20.9 GB98 tok/s8K ctx
dense
Gemma 3 1B1B15.7 GB14 tok/s33K ctx
dense
StarCoder2 15B15B26.7 GB48 tok/s16K ctx
dense
Zephyr 7B Beta7B20.9 GB98 tok/s33K ctx
dense
Starling LM 7B7B20.9 GB98 tok/s8K ctx
dense
Meta Llama 3.1 8B Instruct
C478B20.5 GB95 tok/s1238K ctx
dense
Qwen3.5 9B9B21.3 GB85 tok/s1091K ctx
dense
Qwen3.5 35B A3B35B40.2 GB22 tok/s219K ctx
dense
llava llama 3 8b v1 1
C478B20.5 GB95 tok/s1238K ctx
dense
Qwen3.5 9B Uncensored HauhauCS Aggressive
C479B21.3 GB85 tok/s1091K ctx
dense
DeepSeek R1 0528 Qwen3 8B8B20.5 GB95 tok/s1238K ctx
dense
Meta Llama 3 8B Instruct
C478B20.5 GB95 tok/s1238K ctx
dense
DeepSeek R1 Distill Llama 8B8B20.5 GB95 tok/s1238K ctx
dense
Dolphin3.0 Llama3.1 8B
C478B20.5 GB95 tok/s1238K ctx
dense
Gemma 2 2B2B17.5 GB28 tok/s8K ctx
dense
Llama 3 8B Instruct 32k v0.1
C478B20.5 GB95 tok/s1238K ctx
dense
Qwen3.5 9B
C479B21.3 GB85 tok/s1091K ctx
dense
Meta Llama 3.1 8B Instruct
C478B20.5 GB95 tok/s1238K ctx
dense
vntl llama3 8b v2
C478B20.5 GB95 tok/s1238K ctx
dense
Llama 2 7B Chat
C477B19.8 GB98 tok/s1427K ctx
dense
Qwen3.5 35B A3B
C4735B40.2 GB22 tok/s219K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C478B20.5 GB95 tok/s1238K ctx
dense
Mistral 7B Instruct v0.2
C477B19.8 GB98 tok/s1427K ctx
dense
DeepSeek R1 0528 Qwen3 8B
C478B20.5 GB95 tok/s1238K ctx
dense
Mistral 7B Instruct v0.3
C477B19.8 GB98 tok/s1427K ctx
dense
Qwen3 8B DeepSeek v3.2 Speciale Distill
C478B20.5 GB95 tok/s1238K ctx
dense
Hermes 3 Llama 3.1 8B
C478B20.5 GB95 tok/s1238K ctx
dense
Yi Coder 9B Chat
C479B21.3 GB85 tok/s1091K ctx
dense
glm 4 9b chat 1m
C479B21.3 GB85 tok/s1091K ctx
dense
Hermes 2 Pro Llama 3 8B
C478B20.5 GB95 tok/s1238K ctx
dense
Dolphin3.0 Llama3.1 8B
C478B20.5 GB95 tok/s1238K ctx
dense
granite 8b code instruct 4k
C478B20.5 GB95 tok/s1238K ctx
dense
dolphin 2.9.4 llama3.1 8b
C478B20.5 GB95 tok/s1238K ctx
dense
Yi 1.5 9B Chat
C469B21.3 GB85 tok/s1091K ctx
dense
Hermes 4.3 36B
C4636B40.9 GB21 tok/s210K ctx
dense
Mistral 7B Instruct v0.3
C467B19.8 GB98 tok/s1427K ctx
dense
Hermes 2 Pro Mistral 7B
C467B19.8 GB98 tok/s1427K ctx
dense
openchat 3.6 8b 20240522 IMat
C468B20.5 GB95 tok/s1238K ctx
dense
Qwen3.5 27B27B34.4 GB28 tok/s308K ctx
dense
aya expanse 8b
C468B20.5 GB95 tok/s1238K ctx
dense
zephyr 7B beta
C467B19.8 GB98 tok/s1427K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B20.4 GB98 tok/s1272K ctx
dense
Nous Hermes 2 Mistral 7B DPO
C467B19.8 GB98 tok/s1427K ctx
dense
dolphin v2 8b abliterated i1
C468B20.5 GB95 tok/s1238K ctx
dense
EXAONE 3.5 7.8B Instruct
C467.8B20.4 GB98 tok/s1272K ctx
dense
Falcon H1R 7B
C467B19.8 GB98 tok/s1427K ctx
dense
EXAONE 3.5 7.8B Instruct i1
C467.8B20.4 GB98 tok/s1272K ctx
dense
DeepSeek LLM 7B7B26.3 GB98 tok/s4K ctx
dense
aya expanse 8b orthogonal heretic i1
C468B20.5 GB95 tok/s1238K ctx
dense
falcon mamba 7b instruct Q4 K M
C467B19.8 GB98 tok/s1427K ctx
dense
exaone 3.0 7.8b it
C467.8B20.4 GB98 tok/s1272K ctx
dense
gemma 3 27b it27B34.4 GB28 tok/s308K ctx
dense
zephyr 7B alpha
C467B19.8 GB98 tok/s1427K ctx
dense
speechless zephyr code functionary 7b
C467B19.8 GB98 tok/s1427K ctx
dense
StarCoder2 7B
C467B19.8 GB98 tok/s1427K ctx
dense
gemma 3 12b it
C4612B23.5 GB63 tok/s798K ctx
dense
aya expanse 8b orthogonal heretic
C468B20.5 GB95 tok/s1238K ctx
dense
japanese stablelm instruct gamma 7B
C467B19.8 GB98 tok/s1427K ctx
dense
SOLAR 10.7B Instruct v1.0 uncensored
C4610.7B22.5 GB71 tok/s905K ctx
dense
Qwen 2.5 Coder 0.5B0.5B15.2 GB7 tok/s131K ctx
dense
Yi 9B Coder i1
C469B21.3 GB85 tok/s1091K ctx
dense
Falcon H1 7B Instruct
C467B19.8 GB98 tok/s1427K ctx
dense
Nous Hermes 2 SOLAR 10.7B
C4610.7B22.5 GB71 tok/s905K ctx
dense
Falcon H1R 7B7B19.8 GB98 tok/s1427K ctx
dense
Qwen3.5 122B A10B122B88.8 GB7 tok/s20K ctx
dense
HelpingAI2 9B
C469B21.3 GB85 tok/s1091K ctx
dense
internlm2 5 1 8b chat i1
C468B20.5 GB95 tok/s1238K ctx
dense
internlm3 8b instruct abliterated i1
C468B20.5 GB95 tok/s1238K ctx
dense
EXAONE 4.0 32B
C4632B38.0 GB24 tok/s247K ctx
dense
gemma 3 27b it
C4627B34.4 GB28 tok/s308K ctx
dense
Yi 1.5 6B Chat
C466B19.1 GB84 tok/s1679K ctx
dense
Mamba Codestral 7B v0.1
C467B19.8 GB98 tok/s1427K ctx
dense
aya expanse 32b heretic MPOA i1
C4632B38.0 GB24 tok/s247K ctx
dense
HelpingAI2 9B i1
C469B21.3 GB85 tok/s1091K ctx
dense
stabilityai japanese stablelm base gamma 7b
C467B19.8 GB98 tok/s1427K ctx
dense
DeepSeek R1 Distill Qwen 14B14B24.9 GB54 tok/s672K ctx
dense
HelpingAI 9B 200k i1
C469B21.3 GB85 tok/s1091K ctx
dense
baichuan2 7b chat
C467B19.8 GB98 tok/s1427K ctx
dense
Mistral Small 3.2 24B Instruct 250624B32.2 GB32 tok/s357K ctx
dense
DiscoPOP zephyr 7b gemma
C467B19.8 GB98 tok/s1427K ctx
dense
HelpingAI 9B i1
C469B21.3 GB85 tok/s1091K ctx
dense
stablelm 2 zephyr 1 6b
C466B19.1 GB84 tok/s1679K ctx
dense
starcoder2 7b
C467B19.8 GB98 tok/s1427K ctx
dense
zephyr 7b beta Mistral 7B Instruct v0.2
C467B19.8 GB98 tok/s1427K ctx
dense
OpenChat 3.5 7B Qwen v2.0 i1
C467B19.8 GB98 tok/s1427K ctx
dense
Baichuan M2 32B Q4 K M
C4632B38.0 GB24 tok/s247K ctx
dense
OpenChat 3.5 7B Starling v2.0 i1
C467B19.8 GB98 tok/s1427K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B32.2 GB32 tok/s357K ctx
dense
internlm2 math plus 7b IMat
C467B19.8 GB98 tok/s1427K ctx
dense
Mistral Small 24B Instruct 2501
C4624B32.2 GB32 tok/s357K ctx
dense
CodeNinja 1.0 OpenChat 7B i1
C467B19.8 GB98 tok/s1427K ctx
dense
Neural Chat 7B7B20.9 GB98 tok/s8K ctx
dense
internlm2 5 7b chat i1
C467B19.8 GB98 tok/s1427K ctx
dense
OpenSafetyLab MD Judge v0 2 internlm2 7b
C467B19.8 GB98 tok/s1427K ctx
dense
MD Judge v0 2 internlm2 7b i1
C467B19.8 GB98 tok/s1427K ctx
dense
baichuan inc Baichuan M2 32B
C4632B38.0 GB24 tok/s247K ctx
dense
solar finalised finetuned Model 10.7B i1
C4610.7B22.5 GB71 tok/s905K ctx
dense
zephyr 7b gemma sft african ultrachat 100k
C467B19.8 GB98 tok/s1427K ctx
dense
jointpreferences mistral 7b sft helpful
C467B19.8 GB98 tok/s1427K ctx
dense
zephyr 7b dpo full i1
C467B19.8 GB98 tok/s1427K ctx
dense
blossom v3 baichuan2 7b i1
C467B19.8 GB98 tok/s1427K ctx
dense
Yi 1.5 6B6B19.4 GB84 tok/s4K ctx
dense
mistral small 3.1 24b instruct 2503 hf
C4624B32.2 GB32 tok/s357K ctx
dense
blossom v1 baichuan 7b i1
C467B19.8 GB98 tok/s1427K ctx
dense
cognitivecomputations Dolphin3.0 R1 Mistral 24B
C4624B32.2 GB32 tok/s357K ctx
dense
SOLAR 10.7B v1.0
C4610.7B22.5 GB71 tok/s905K ctx
dense
HelpingAI2.5 10B i1
C4610B22.0 GB76 tok/s974K ctx
dense
Yi 1.5 6B Chat
C466B19.1 GB84 tok/s1679K ctx
dense
stablelm 2 1 6b chat imatrix
C466B19.1 GB84 tok/s1679K ctx
dense
NousResearch Hermes 4 14B
C4614B24.9 GB54 tok/s672K ctx
dense
Dolphin Mistral GLM 4.7 Flash 24B Venice Edition Thinking Uncensored i1
C4624B32.2 GB32 tok/s357K ctx
dense
Phi 4 reasoning vision 15B
C4615B25.6 GB51 tok/s622K ctx
dense
Helply 10.2b chat i1
C4610.2B22.1 GB75 tok/s953K ctx
dense
Codestral 22B v0.1
C4622B30.7 GB35 tok/s397K ctx
dense
StarCoder2 15B
C4615B25.6 GB51 tok/s622K ctx
dense
Solar Open 100B
C46100B87.4 GB8 tok/s22K ctx
dense
Solar Open 100B i1
C46100B87.4 GB8 tok/s22K ctx
dense
Codestral 22B v0.1
C4622B30.7 GB35 tok/s397K ctx
dense
GGUF SOLARized GraniStral 14B 2102 YeAM HCT 32QKV
C4614B24.9 GB54 tok/s672K ctx
dense
Codestral 22B v0.1
C4622B30.7 GB35 tok/s397K ctx
dense
StarCoder2 7B7B19.5 GB98 tok/s16K ctx
dense
HelpingAI2 6B
C466B19.1 GB84 tok/s1679K ctx
dense
starcoder2 15b instruct v0.1
C4615B25.6 GB51 tok/s622K ctx
dense
internlm JanusCoder 14B
C4614B24.9 GB54 tok/s672K ctx
dense
starcoder2 15b i1
C4615B25.6 GB51 tok/s622K ctx
dense
GGUF SOLARized GraniStral 14B 1902 YeAM HCT
C4614B24.9 GB54 tok/s672K ctx
dense
cognitivecomputations Dolphin Mistral 24B Venice Edition
C4624B32.2 GB32 tok/s357K ctx
dense
HelpingAI2 6B i1
C466B19.1 GB84 tok/s1679K ctx
dense
starcoder2 15b instruct v0.1
C4615B25.6 GB51 tok/s622K ctx
dense
Nemotron Mini 4B4B19.1 GB56 tok/s4K ctx
dense
Codestral 22B v0.1 IMat
C4522B30.7 GB35 tok/s397K ctx
dense
HelpingAI 15B i1
C4515B25.6 GB51 tok/s622K ctx
dense
internlm2 5 20b chat
C4520B29.3 GB38 tok/s445K ctx
dense
internlm2 limarp chat 20b
C4520B29.3 GB38 tok/s445K ctx
dense
Codestral 21B Pruned i1
C4521B30.0 GB36 tok/s420K ctx
dense
internlm2 math plus 20b i1
C4520B29.3 GB38 tok/s445K ctx
dense
Codestral 22B v0.1 i1
C4522B30.7 GB35 tok/s397K ctx
dense
Codestral RAG 19B Pruned i1
C4519B28.5 GB40 tok/s473K ctx
dense
HelpingAI2.5 5B i1
C455B18.4 GB70 tok/s2031K ctx
dense
Qwen3.5 4B4B17.6 GB56 tok/s2560K ctx
dense
gemma 3 4b it
C454B17.6 GB56 tok/s2560K ctx
dense
gemma 3 4b it
C444B17.6 GB56 tok/s2560K ctx
dense
Qwen 3 0.6B0.6B15.9 GB8 tok/s33K ctx
dense
Llama 3.2 1B1B15.8 GB14 tok/s128K ctx
dense
Llama 3.2 3B Instruct
C443B17.2 GB42 tok/s3426K ctx
dense
Qwen2.5 3B Instruct3B16.9 GB42 tok/s3441K ctx
dense
Llama 3.2 3B Instruct
C433B16.9 GB42 tok/s3441K ctx
dense
Ministral 3 3B Instruct 2512
C433B16.9 GB42 tok/s3441K ctx
dense
HELVETE 3B
C433B16.9 GB42 tok/s3441K ctx
dense
Hermes 3 Llama 3.2 3B
C433B16.9 GB42 tok/s3441K ctx
dense
stablelm zephyr 3b
C433B16.9 GB42 tok/s3441K ctx
dense
StarCoder2 3B
C433B16.9 GB42 tok/s3441K ctx
dense
AI21 Jamba Reasoning 3B
C433B16.9 GB42 tok/s3441K ctx
dense
stablelm 3b 4e1t
C433B16.9 GB42 tok/s3441K ctx
dense
ai21labs AI21 Jamba Reasoning 3B
C433B16.9 GB42 tok/s3441K ctx
dense
ai21labs AI21 Jamba2 3B
C433B16.9 GB42 tok/s3441K ctx
dense
HelpingAI 3B hindi i1
C433B16.9 GB42 tok/s3441K ctx
dense
AI21 Jamba2 3B
C433B16.9 GB42 tok/s3441K ctx
dense
HelpingAI 3B hindi
C433B16.9 GB42 tok/s3441K ctx
dense
AI21 Jamba2 3B i1
C433B16.9 GB42 tok/s3441K ctx
dense
EXAONE 3.5 2.4B Instruct
C422.4B16.5 GB34 tok/s4322K ctx
dense
gemma 2 2b it
C422B16.6 GB28 tok/s5174K ctx
dense
gemma 2b2B16.2 GB28 tok/s5203K ctx
dense
Gemmasutra Mini 2B v1
C422B16.2 GB28 tok/s5203K ctx
dense
gemma 2 2b it
C422B16.2 GB28 tok/s5203K ctx
dense
Qwen 3.5 0.6B0.6B15.9 GB8 tok/s131K ctx
dense
StarCoder2 3B3B17.0 GB42 tok/s16K ctx
dense
Qwen2.5 1.5B Instruct1.5B15.8 GB21 tok/s6965K ctx
dense
DeepSeek R1 Distill Qwen 1.5B1.5B15.8 GB21 tok/s6965K ctx
dense
Yi Coder 1.5B Chat
C411.5B15.8 GB21 tok/s6965K ctx
dense
stablelm 2 zephyr 1.6b
C411.6B15.9 GB22 tok/s6525K ctx
dense
logos16v2 stablelm2 1.6b i1
C411.6B15.9 GB22 tok/s6525K ctx
dense
Yi Coder 1.5B
C411.5B15.8 GB21 tok/s6965K ctx
dense
Falcon H1 1.5B Instruct1.5B15.8 GB21 tok/s6965K ctx
dense
TinyLlama 1.1B Chat v1.0
C401.1B15.5 GB15 tok/s9528K ctx
dense
F
F Tier -- Can't Run (36 models)
Too heavy for this hardware.
Qwen 2.5 0.5B0.5B15.2 GB7 tok/s131K ctx
dense
EXAONE 4.0 1.2B
D401.2B15.6 GB17 tok/s8727K ctx
dense
Llama 3.2 1B Instruct Q8 0
D401B15.7 GB14 tok/s10461K ctx
dense
TinyLlama 1.1B Chat v0.3
D401.1B15.5 GB15 tok/s9528K ctx
dense
TinyLlama 1.1B Chat v0.61.1B15.5 GB15 tok/s9528K ctx
dense
Llama 3.2 1B Instruct
D401B15.5 GB14 tok/s10489K ctx
dense
gemma 3 1b it
D401B15.5 GB14 tok/s10489K ctx
dense
TinyLlama 1.1B Chat v1.0 imatrix
D391.1B15.5 GB15 tok/s9528K ctx
dense
Falcon3 1B Instruct abliterated
D391B15.5 GB14 tok/s10489K ctx
dense
SmolVLM 500M Instruct
D380.5B15.2 GB7 tok/s12324K ctx
dense
embeddinggemma 300M
D360.3B15.1 GB4 tok/s12350K ctx
dense
granite embedding 107m multilingual
D340.11B14.9 GB2 tok/s12379K ctx
dense
Falcon H1 Tiny 90M Instruct
D340.09B14.9 GB2 tok/s12381K ctx
dense
Qwen 3 235B A22B235B160.9 GB5 tok/s4K ctx
moe
Qwen3-Coder 480B A35B Instruct480B310.4 GB3 tok/s4K ctx
moe
Qwen 3.5 397B A17B397B259.8 GB4 tok/s4K ctx
moe
Mistral Large 3675B431.8 GB3 tok/s4K ctx
moe
GLM-5744B487.6 GB2 tok/s4K ctx
moe
GLM-5.1754B493.7 GB2 tok/s4K ctx
moe
Kimi K2.51000B632.2 GB2 tok/s4K ctx
moe
Kimi K2.61000B632.2 GB2 tok/s4K ctx
moe
DeepSeek V3 671B671B483.6 GB2 tok/s4K ctx
moe
DeepSeek V3.1 671B671B483.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Pro1600B878.6 GB2 tok/s4K ctx
moe
DeepSeek V4 Flash284B174.0 GB7 tok/s4K ctx
moe
DeepSeek R1 671B671B483.6 GB2 tok/s4K ctx
moe
DeepSeek Coder V2 236B236B217.3 GB6 tok/s4K ctx
moe
DeepSeek V2.5 236B236B217.3 GB6 tok/s4K ctx
moe
Llama 4 Maverick 17B 128E400B261.7 GB4 tok/s4K ctx
moe
Llama 3.1 405B405B269.5 GB2 tok/s4K ctx
dense
DeepSeek V3.2671B424.5 GB3 tok/s4K ctx
moe
MiniMax M2.7
F0230B158.8 GB8 tok/s4K ctx
moe
K EXAONE 236B A23B
F0236B186.3 GB2 tok/s4K ctx
dense
Baichuan M3 235B
F0235B185.6 GB2 tok/s4K ctx
dense
Baichuan M3 235B i1
F0235B185.6 GB2 tok/s4K ctx
dense
Qwen3.5 397B A17B397B303.4 GB2 tok/s4K ctx
dense
Compare with...
See How Other Hardware Stacks Up
Check tier lists for similar GPUs and Apple Silicon configs