RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the Mac M1 16GB?

The Mac M1 16GB has 16 GB of unified memory (68 GB/s). About 10.7 GB is usable for a model. 165 of 285 open models we track fit at 8K context.

165models fit (8K ctx)
144still fit at 32K ctx
26Blargest that fits
68 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
Trinity Mini26B (2.0B act.)Q2_K9.8 GBQ2_K96details
GLiDE 27B26BQ2_K10.5 GBno7.0details
diffusiongemma 26B A4B it26B (4.0B act.)Q2_K10.4 GBno45details
gemma 4 26B A4B it26B (4.0B act.)Q2_K10.4 GBno45details
LFM2.5 24B A2B24B (2.0B act.)Q2_K9.2 GBQ2_K91details
ERNIE 4.5 21B A3B PT22BQ2_K8.6 GBQ2_K8.3details
ERNIE 4.5 21B A3B Thinking22BQ2_K8.6 GBQ2_K8.3details
Groot 19B A3B 26102319B (3.0B act.)Q2_K8.4 GBQ2_K56details
Kimi VL A3B Thinking 250616B (3.0B act.)Q3_K_M8.4 GBQ3_K_M46details
Kimi VL A3B Instruct16B (3.0B act.)Q3_K_M8.7 GBQ3_K_M45details
Kimi VL A3B Thinking16B (3.0B act.)Q3_K_M8.4 GBQ3_K_M46details
deepseek moe 16b base16B (2.8B act.)Q2_K8.6 GBno59details
LLaDA2.1 mini16B (1.4B act.)Q3_K_M8.4 GBQ3_K_M97details
Moonlight 16B A3B Instruct16B (3.0B act.)Q3_K_M8.7 GBQ3_K_M44details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q3_K_M8.0 GBQ3_K_M52details
DeepSeek V2 Lite16B (2.7B act.)Q3_K_M8.0 GBQ3_K_M52details
DeepSeek V2 Lite Chat16B (2.7B act.)Q3_K_M8.4 GBQ3_K_M49details
Qwen2.5 Coder 14B Instruct15BQ3_K_M8.9 GBno9.4details
DeepSeek R1 Distill Qwen 14B15BQ3_K_M9.0 GBno9.3details
Qwen3 14B15BQ3_K_M8.7 GBno9.4details
phi 415BQ3_K_M8.9 GBno9.5details
gemma 3 12b it12BQ5_K_M9.6 GBQ3_K_M8.1details
Llama Guard 4 12B12BQ4_K_M9.1 GBno9.3details
Llama 3.2 11B Vision Instruct11BQ5_K_M9.0 GBQ2_K8.9details
Llama 3.2 11B Vision11BQ5_K_M9.0 GBQ2_K9.0details
GLM 4.6V Flash10BQ6_K8.7 GBQ6_K8.2details
GLM 4.1V 9B Thinking10BQ6_K8.7 GBQ6_K8.2details
Step3 VL 10B10BQ6_K8.1 GBQ3_K_M10details
SenseNova Loop 9B Preview9.7BQ6_K8.0 GBQ6_K9.2details
Qwen3.5 9B9.7BQ6_K8.1 GBQ6_K9.1details
Ornith 1.5 9B9.7BQ6_K8.3 GBQ6_K8.8details
Qwen3.5 9B Base9.7BQ6_K8.0 GBQ6_K9.2details
MiMo V2.6 Distill Qwen 9B9.4BQ6_K8.4 GBQ6_K8.7details
glm 4 9b9.4BQ2_K9.4 GBno17details
internlm3 8b instruct8.8BQ6_K7.8 GBQ6_K9.4details
granite 4.2 8b8.8BQ6_K8.6 GBQ3_K_M9.4details
granite 4.1 8b8.8BQ6_K8.6 GBQ3_K_M9.4details
Qwen3 VL 8B Instruct8.8BQ6_K8.5 GBQ3_K_M9.5details
Molmo2 8B8.7BQ6_K8.4 GBQ3_K_M9.6details
LFM2.5 8B A1B8.5B (1.0B act.)Q8_09.1 GBQ8_064details
Qwen2.5 VL 7B Instruct8.3BQ8_08.7 GBQ6_K8.4details
UI TARS 1.5 7B8.3BQ8_08.7 GBQ6_K8.4details
olmOCR 2 7B 10258.3BQ8_08.7 GBQ6_K8.4details
olmOCR 7B 0225 preview8.3BQ8_08.7 GBQ6_K8.4details
UI TARS 7B DPO8.3BQ8_08.7 GBQ6_K8.4details
Qwen3 8B8.2BQ6_K8.1 GBQ3_K_M10details
DeepSeek R1 0528 Qwen3 8B8.2BQ6_K8.1 GBQ3_K_M10details
granite 3.3 8b instruct8.2BQ6_K8.0 GBQ3_K_M10details
granite 3.1 8b instruct8.2BQ6_K8.0 GBQ3_K_M10details
granite 3.2 8b instruct8.2BQ6_K8.0 GBQ3_K_M10details
Apertus 8B Instruct 25098.1BQ8_09.6 GBQ4_K_M7.9details
Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ4_K_M8.0details
Llama 3.1 8B8.0BQ8_09.6 GBQ4_K_M8.0details
Hermes 3 Llama 3.1 8B8.0BQ8_09.6 GBQ4_K_M8.0details
DeepSeek R1 Distill Llama 8B8.0BQ8_09.6 GBQ4_K_M8.0details
Meta Llama 3 8B Instruct8.0BQ8_09.6 GBQ4_K_M8.0details
Meta Llama Guard 2 8B8.0BQ8_09.6 GBQ4_K_M8.0details
Meta Llama 3.1 8B Instruct8.0BQ8_09.6 GBQ4_K_M8.0details
Ling 3.0 tiny7.9B (1.4B act.)Q8_08.7 GBQ8_046details
MiMo 7B RL7.8BQ8_09.4 GBQ4_K_M8.4details
internlm2 5 7b chat7.7BQ8_09.2 GBQ4_K_M8.3details
internlm2 chat 7b7.7BQ8_09.2 GBQ4_K_M8.3details
Qwen2.5 7B Instruct7.6BQ8_08.7 GBQ6_K8.4details
Qwen2.5 Coder 7B Instruct7.6BQ8_08.7 GBQ6_K8.4details
DeepSeek R1 Distill Qwen 7B7.6BQ8_08.7 GBQ6_K8.4details
Falcon H1R 7B7.6BQ8_08.5 GBQ8_08.4details
Olmo 3 7B Instruct7.3BQ6_K8.7 GBQ3_K_M11details
Olmo 3 1025 7B7.3BQ6_K8.7 GBQ3_K_M11details
Olmo 3 7B Think7.3BQ6_K8.7 GBQ3_K_M11details
Mistral 7B v0.17.2BQ8_08.6 GBQ5_K_M8.8details
Mistral 7B Instruct v0.27.2BQ8_08.6 GBQ5_K_M8.8details
Mistral 7B Instruct v0.17.2BQ8_08.6 GBQ5_K_M8.8details
falcon 7b7.2BQ4_K_M9.6 GBno14details
falcon 7b instruct7.2BQ4_K_M9.4 GBno14details
granite 4.0 h tiny6.9B (1.5B act.)Q8_07.5 GBQ8_044details
OLMoE 1B 7B 09246.9B (1.3B act.)Q8_08.4 GBQ5_K_M50details
deepseek coder 7b instruct v1.56.9BQ6_K9.6 GBno12details
deepseek coder 6.7b instruct6.7BQ6_K9.6 GBno12details
deepseek coder 6.7b base6.7BQ6_K9.6 GBno12details
Llama 2 7b hf6.7BQ6_K9.6 GBno12details
Yi 6B 200K6.1BQ8_07.0 GBQ8_011details
Yi 6B6.1BQ8_07.0 GBQ8_011details
Molmo2 4B4.9BQ8_06.2 GBQ8_014details
WeVisOne VL 4B4.8BQ8_06.2 GBQ8_014details
Mage VL Cowork4.7BQ8_06.2 GBQ8_014details
Qianfan OCR4.7BQ8_06.2 GBQ8_014details
Qwen3.5 4B4.7BF169.0 GBQ8_08.1details
S5 nano4.7BF169.2 GBQ8_07.8details
LightOnOCR 3 4B4.5BF169.0 GBQ8_08.1details
SingVL S1 4B4.5BF169.0 GBQ8_08.1details
Intern Decision 4B4.5BF169.0 GBQ8_08.1details
Qwen3 VL 4B Instruct4.4BF169.3 GBQ8_08.4details
gemma 3 4b it4.3BF168.4 GBF168.8details
medgemma 1.5 4b it4.3BF168.4 GBF168.8details
medgemma 4b it4.3BF168.4 GBF168.8details
gemma 3 4b pt4.3BF169.2 GBF167.9details
Phi 3.5 vision instruct4.1BQ8_07.6 GBno15details
Qwen3 4B4.0BF169.3 GBQ8_08.5details
Qwen3 4B Instruct 25074.0BF169.3 GBQ8_08.4details
granite vision 4.1 4b4.0BF167.6 GBF1610.0details
Phi 4 mini instruct3.8BF168.9 GBQ8_08.9details
Phi 3 mini 4k instruct3.8BQ8_07.2 GBno17details
Phi 3.5 mini instruct3.8BQ8_07.2 GBno17details
Phi 3 mini 128k instruct3.8BQ8_07.2 GBno17details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF1611details
granite 4.2 3b3.7BF168.0 GBQ8_09.3details
granite 4.1 3b llmon preview3.7BF168.0 GBQ8_09.3details
Raon 1 nano3.5BF167.0 GBF1611details
granite 4.1 3b3.4BF167.6 GBF1610.0details
granite 4.0 micro3.4BF167.6 GBF1610.0details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF1633details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF1629details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF1634details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF1634details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF1634details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF1638details
Llama 3.2 3B Instruct3.2BQ8_04.7 GBQ8_020details
Llama 3.2 3B3.2BF167.5 GBQ8_011details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBQ8_011details
d1 3B3.1BQ8_03.4 GBQ8_024details
LFM2.5 VL 3B3.1BQ8_03.4 GBQ8_024details
Qwen2.5 3B Instruct3.1BF166.7 GBF1611details
SmolLM3 3B3.1BF166.9 GBF1611details
SmolLM3 3B Base3.1BF166.9 GBF1611details
phi 22.8BF168.2 GBno12details
LFM2.5 2.6B2.7BQ8_03.4 GBQ8_024details
granite 3.3 2b instruct2.5BF165.8 GBF1613details
MiniCPM5 2B2.5BF165.7 GBF1613details
WeVisOne VL 2B2.4BF165.4 GBF1617details
Qwen3.5 2B2.3BF164.5 GBF1618details
SmolVLM Instruct2.2BF165.4 GBQ8_019details
Qwen3 1.7B2.0BF164.8 GBF1620details
Qwen3.8 27B DSpine b162.0BF164.8 GBF1617details
d1 2B1.9BF167.2 GBF169.8details
LFM3 VL 2B1.9BF167.2 GBF169.8details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF1619details
LFM3 1.7B1.7BF167.2 GBF169.8details
LFM3 1.7B Base1.7BF164.0 GBF1620details
SmolLM2 1.7B Instruct1.7BF165.2 GBQ8_020details
SmolLM2 1.7B1.7BF165.2 GBQ8_020details
SmolLM 1.7B1.7BF165.2 GBQ8_020details
granite 4.0 1b1.6BF164.3 GBF1621details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF1622details
OLMo 2 0425 1B1.5BF164.4 GBF1623details
deepseek coder 1.3b instruct1.3BF164.5 GBF1625details
Llama 3.2 1B1.2BF163.2 GBF1628details
Llama 3.2 1B Instruct1.2BF165.5 GBF1614details
LFM2.5 1.2B Instruct1.2BF165.0 GBF1615details
HunyuanOCR1.1BF163.5 GBF1630details
PaddleOCR VL 1.6959MF162.5 GBF1635details
Qwen3.5 0.8B873MF162.4 GBF1645details
LightOnOCR 3 0.8B853MF162.4 GBF1645details
FastVLM 0.5B759MF162.2 GBF1645details
Qwen3 0.6B752MF162.7 GBF1657details
GOT OCR2 0716MF162.8 GBF1647details
Nemotron 3.5 Parse631MF161.9 GBF1652details
GOT OCR 2.0 hf561MF162.5 GBF1661details
SmolVLM2 500M Video Instruct507MF161.6 GBF1683details
SmolVLM 500M Instruct507MF161.6 GBF1683details
Qwen2.5 0.5B Instruct494MF161.7 GBF1669details
LFM2.5 VL 450M449MF161.9 GBF1648details
SmolLM2 360M Instruct362MF161.5 GBF1694details
SmolLM2 360M362MF161.5 GBF1694details
LFM2.5 350M354MF161.9 GBF1648details
LFM2 350M354MF161.4 GBF1696details

Just too big at 4-bit

These need up to 35% more memory than the Mac M1 16GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Moonlight 16B A3B Instruct10.8 GB0.0 GB
Kimi VL A3B Thinking 250610.8 GB0.0 GB
Kimi VL A3B Instruct10.8 GB0.0 GB
glm 4 9b11.5 GB0.8 GB
gpt oss safeguard 20b11.8 GB1.1 GB
gpt oss 20b11.8 GB1.1 GB
deepseek moe 16b base12.5 GB1.7 GB
Groot 19B A3B 26102312.7 GB1.9 GB
ERNIE 4.5 21B A3B PT13.5 GB2.7 GB
ERNIE 4.5 21B A3B Thinking13.5 GB2.7 GB
LFM2.5 24B A2B14.3 GB3.6 GB

Questions

What is the biggest model that runs on the Mac M1 16GB?

Of the models we track, the largest that fits at 8K context is Trinity Mini (26B) at Q2_K, needing about 9.8 GB.

How much of the Mac M1 16GB's memory can a model use?

About 11 GB of 16 GB. macOS lets the GPU use about 67% of unified memory by default; this can be raised with the iogpu.wired_limit_mb setting.

Spec source: Apple specifications. Model data: Hugging Face Hub API. Speed is a bandwidth ceiling (bandwidth ÷ bytes read per token), not a benchmark. Methodology · Report an error.