RunsOnMy ModelsGPUs & MacsAPI pricesVRAM calculatorNew

What AI models run on the Mac M4 32GB?

The Mac M4 32GB has 32 GB of unified memory (120 GB/s). About 21.4 GB is usable for a model. 202 of 285 open models we track fit at 8K context.

202models fit (8K ctx)
196still fit at 32K ctx
49Blargest that fits
120 GB/smemory bandwidth

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
Kimi Linear 48B A3B Instruct49B (3.0B act.)Q2_K17.3 GBQ2_K112details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)Q2_K17.6 GBQ2_K25details
falcon 40b42BQ2_K16.1 GBQ2_K7.6details
flint40B (1.3B act.)Q3_K_M19.0 GBQ2_K196details
Maxwell V1 Flash37B (2.3B act.)Q3_K_M17.9 GBQ3_K_M108details
Hermes 4.3 36B36BQ3_K_M19.1 GBQ2_K6.8details
Seed OSS 36B Instruct36BQ3_K_M19.1 GBQ2_K6.8details
Qwen3.6 35B A3B36B (3.0B act.)Q3_K_M17.0 GBQ3_K_M84details
Ornith 1.5 35B A3B36B (3.0B act.)Q3_K_M16.6 GBQ3_K_M86details
Qwen3.5 35B A3B36B (3.0B act.)Q3_K_M16.3 GBQ3_K_M88details
Axion 1.0 35B35B (3.9B act.)Q3_K_M16.6 GBQ3_K_M64details
Qwen AgentWorld 35B A3B35B (3.0B act.)Q3_K_M16.6 GBQ3_K_M83details
Yi 34B 200K34BQ3_K_M17.9 GBno7.2details
Yi 34B34BQ3_K_M17.9 GBno7.2details
Yi 34B Chat34BQ3_K_M17.9 GBno7.2details
Yi 1.5 34B Chat34BQ3_K_M18.1 GBQ2_K7.1details
Nous Hermes 2 Yi 34B34BQ3_K_M18.4 GBQ2_K7.0details
DeepSeek R1 Distill Qwen 32B33BQ3_K_M17.5 GBQ2_K7.5details
Qwen3 32B33BQ3_K_M17.6 GBQ2_K7.5details
GLM Z1 32B 041433BQ3_K_M16.0 GBQ3_K_M7.6details
GLM 4 32B 041433BQ3_K_M16.0 GBQ3_K_M7.6details
granite 4.0 h small32B (8.8B act.)Q4_K_M18.9 GBQ4_K_M23details
gemma 4 31B it31BQ3_K_M16.7 GBQ2_K8.1details
GLM 4.7 Flash31B (4.6B act.)Q4_K_M18.2 GBQ3_K_M44details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)Q4_K_M18.7 GBQ3_K_M66details
North Mini Code 1.030B (3.3B act.)Q4_K_M19.3 GBQ3_K_M58details
Hy Memory 30B A3B30B (3.0B act.)Q4_K_M18.5 GBQ3_K_M65details
Muse Glimmer 30B30BQ5_K_M18.8 GBQ5_K_M6.3details
granite 4.2 30b29BQ4_K_M19.1 GBQ2_K6.8details
Qwen3.8 27B28BQ4_K_M16.8 GBQ4_K_M7.2details
Qwen3.6 27B28BQ4_K_M17.0 GBQ4_K_M7.1details
Qwen3.5 27B28BQ4_K_M17.0 GBQ4_K_M7.2details
gemma 3 27b it27BQ4_K_M17.3 GBQ4_K_M7.3details
Arete 27B27BQ4_K_M17.4 GBQ4_K_M7.0details
MiniMax H3 Image PE27BQ5_K_M19.3 GBQ4_K_M6.2details
Trinity Mini26B (2.0B act.)Q5_K_M18.4 GBQ5_K_M86details
GLiDE 27B26BQ5_K_M18.7 GBQ4_K_M6.5details
diffusiongemma 26B A4B it26B (4.0B act.)Q5_K_M19.2 GBQ4_K_M40details
gemma 4 26B A4B it26B (4.0B act.)Q4_K_M17.2 GBQ4_K_M46details
LFM2.5 24B A2B24B (2.0B act.)Q6_K19.1 GBQ5_K_M73details
ERNIE 4.5 21B A3B PT22BQ6_K17.8 GBQ6_K6.7details
ERNIE 4.5 21B A3B Thinking22BQ6_K17.8 GBQ6_K6.7details
gpt oss safeguard 20b22B (4.8B act.)F1613.8 GBF1639details
gpt oss 20b21B (4.2B act.)F1613.8 GBF1643details
Groot 19B A3B 26102319B (3.0B act.)Q6_K16.7 GBQ6_K46details
Kimi VL A3B Thinking 250616B (3.0B act.)Q8_016.8 GBQ8_039details
Kimi VL A3B Instruct16B (3.0B act.)Q8_016.8 GBQ8_039details
Kimi VL A3B Thinking16B (3.0B act.)Q8_017.2 GBQ8_038details
deepseek moe 16b base16B (2.8B act.)Q8_018.6 GBQ5_K_M40details
LLaDA2.1 mini16B (1.4B act.)Q8_017.1 GBQ8_079details
Moonlight 16B A3B Instruct16B (3.0B act.)Q8_016.8 GBQ8_038details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)Q8_016.4 GBQ8_042details
DeepSeek V2 Lite16B (2.7B act.)Q8_016.4 GBQ8_042details
DeepSeek V2 Lite Chat16B (2.7B act.)Q8_016.4 GBQ8_042details
Qwen2.5 Coder 14B Instruct15BQ8_016.8 GBQ6_K7.6details
DeepSeek R1 Distill Qwen 14B15BQ8_016.8 GBQ6_K7.6details
Qwen3 14B15BQ8_016.6 GBQ6_K7.6details
phi 415BQ8_016.7 GBQ6_K7.7details
gemma 3 12b it12BQ8_013.4 GBQ8_09.6details
Llama Guard 4 12B12BQ8_014.2 GBQ8_09.4details
Llama 3.2 11B Vision Instruct11BQ8_011.6 GBQ8_012details
Llama 3.2 11B Vision11BQ8_012.4 GBQ8_011details
GLM 4.6V Flash10BF1618.5 GBQ8_06.4details
GLM 4.1V 9B Thinking10BF1618.5 GBQ8_06.4details
Step3 VL 10B10BQ8_09.9 GBQ8_014details
SenseNova Loop 9B Preview9.7BF1617.8 GBF166.7details
Qwen3.5 9B9.7BF1617.8 GBF166.7details
Ornith 1.5 9B9.7BF1617.8 GBF166.7details
Qwen3.5 9B Base9.7BF1617.8 GBF166.7details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF166.7details
glm 4 9b9.4BQ8_015.0 GBno12details
internlm3 8b instruct8.8BF1617.4 GBF166.8details
granite 4.2 8b8.8BF1618.2 GBQ8_06.8details
granite 4.1 8b8.8BF1618.2 GBQ8_06.8details
Qwen3 VL 8B Instruct8.8BF1617.1 GBQ8_07.3details
Molmo2 8B8.7BF1617.1 GBQ8_07.3details
LFM2.5 8B A1B8.5B (1.0B act.)Q8_09.1 GBQ8_0113details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF167.9details
UI TARS 1.5 7B8.3BF1615.3 GBF167.9details
olmOCR 2 7B 10258.3BF1615.3 GBF167.9details
olmOCR 7B 0225 preview8.3BF1616.6 GBF167.2details
UI TARS 7B DPO8.3BF1615.3 GBF167.9details
Qwen3 8B8.2BF1617.1 GBQ8_07.3details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBQ8_07.3details
granite 3.3 8b instruct8.2BF1617.0 GBQ8_07.3details
granite 3.1 8b instruct8.2BF1617.0 GBQ8_07.3details
granite 3.2 8b instruct8.2BF1617.0 GBQ8_07.3details
Apertus 8B Instruct 25098.1BF1616.7 GBQ8_07.4details
Llama 3.1 8B Instruct8.0BF1616.6 GBQ8_07.5details
Llama 3.1 8B8.0BF1616.6 GBQ8_07.5details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBQ8_07.5details
DeepSeek R1 Distill Llama 8B8.0BQ8_09.6 GBQ8_014details
Meta Llama 3 8B Instruct8.0BF1616.6 GBQ8_07.5details
Meta Llama Guard 2 8B8.0BF1616.6 GBQ8_07.5details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBQ8_07.5details
Ling 3.0 tiny7.9B (1.4B act.)F1615.6 GBF1643details
MiMo 7B RL7.8BF1616.4 GBQ8_07.7details
internlm2 5 7b chat7.7BF1616.0 GBF167.8details
internlm2 chat 7b7.7BF1616.0 GBF167.8details
Qwen2.5 7B Instruct7.6BF1615.3 GBF167.9details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF167.9details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF167.9details
Falcon H1R 7B7.6BF1615.1 GBF167.9details
Olmo 3 7B Instruct7.3BF1616.7 GBQ8_08.2details
Olmo 3 1025 7B7.3BF1616.7 GBQ8_08.2details
Olmo 3 7B Think7.3BF1616.7 GBQ8_08.2details
Mistral 7B v0.17.2BF1614.9 GBF168.3details
Mistral 7B Instruct v0.27.2BF1614.9 GBF168.3details
Mistral 7B Instruct v0.17.2BF1614.9 GBF168.3details
falcon 7b7.2BF1618.4 GBno8.3details
falcon 7b instruct7.2BF1618.4 GBno8.3details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBF1641details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF1647details
deepseek coder 7b instruct v1.56.9BF1617.2 GBQ3_K_M8.7details
deepseek coder 6.7b instruct6.7BF1617.0 GBQ2_K8.9details
deepseek coder 6.7b base6.7BF1617.0 GBQ2_K8.9details
Llama 2 7b hf6.7BF1617.0 GBQ2_K8.9details
Yi 6B 200K6.1BF1612.3 GBF169.9details
Yi 6B6.1BF1612.3 GBF169.9details
Molmo2 4B4.9BF1610.0 GBF1614details
WeVisOne VL 4B4.8BF1610.0 GBF1614details
Mage VL Cowork4.7BF1610.0 GBF1614details
Qianfan OCR4.7BF1610.1 GBF1614details
Qwen3.5 4B4.7BF169.0 GBF1614details
S5 nano4.7BF169.2 GBF1614details
LightOnOCR 3 4B4.5BF169.0 GBF1614details
SingVL S1 4B4.5BF169.0 GBF1614details
Intern Decision 4B4.5BF169.0 GBF1614details
Qwen3 VL 4B Instruct4.4BF169.3 GBF1615details
gemma 3 4b it4.3BF168.4 GBF1615details
medgemma 1.5 4b it4.3BF168.4 GBF1615details
medgemma 4b it4.3BF168.4 GBF1615details
gemma 3 4b pt4.3BF169.2 GBF1614details
Phi 3.5 vision instruct4.1BF1611.2 GBQ8_014details
Qwen3 4B4.0BF169.3 GBF1615details
Qwen3 4B Instruct 25074.0BF169.3 GBF1615details
granite vision 4.1 4b4.0BF167.6 GBF1618details
Phi 4 mini instruct3.8BF168.9 GBF1616details
Phi 3 mini 4k instruct3.8BF1610.6 GBQ8_016details
Phi 3.5 mini instruct3.8BF1610.6 GBQ8_016details
Phi 3 mini 128k instruct3.8BF1610.6 GBQ8_016details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF1619details
granite 4.2 3b3.7BF168.0 GBF1616details
granite 4.1 3b llmon preview3.7BF168.0 GBF1616details
Raon 1 nano3.5BF167.0 GBF1619details
granite 4.1 3b3.4BF167.6 GBF1618details
granite 4.0 micro3.4BF167.6 GBF1618details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF1658details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF1651details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF1659details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF1660details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF1660details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF1668details
Llama 3.2 3B Instruct3.2BF1613.5 GBF169.3details
Llama 3.2 3B3.2BF167.5 GBF1619details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF1619details
d1 3B3.1BF1610.8 GBF1611details
LFM2.5 VL 3B3.1BF1610.8 GBF1611details
Qwen2.5 3B Instruct3.1BF166.7 GBF1619details
SmolLM3 3B3.1BF166.9 GBF1619details
SmolLM3 3B Base3.1BF166.9 GBF1620details
phi 22.8BF168.2 GBF1622details
LFM2.5 2.6B2.7BF1610.8 GBF1611details
granite 3.3 2b instruct2.5BF165.8 GBF1624details
MiniCPM5 2B2.5BF165.7 GBF1624details
WeVisOne VL 2B2.4BF165.4 GBF1629details
Qwen3.5 2B2.3BF164.5 GBF1632details
SmolVLM Instruct2.2BF165.4 GBF1633details
Qwen3 1.7B2.0BF164.8 GBF1635details
Qwen3.8 27B DSpine b162.0BF164.8 GBF1630details
d1 2B1.9BF167.2 GBF1617details
LFM3 VL 2B1.9BF167.2 GBF1617details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF1634details
LFM3 1.7B1.7BF167.2 GBF1617details
LFM3 1.7B Base1.7BF164.0 GBF1635details
SmolLM2 1.7B Instruct1.7BF165.2 GBF1635details
SmolLM2 1.7B1.7BF165.2 GBF1635details
SmolLM 1.7B1.7BF165.2 GBF1635details
granite 4.0 1b1.6BF164.3 GBF1637details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF1639details
OLMo 2 0425 1B1.5BF164.4 GBF1640details
deepseek coder 1.3b instruct1.3BF164.5 GBF1645details
Llama 3.2 1B1.2BF163.2 GBF1649details
Llama 3.2 1B Instruct1.2BF165.5 GBF1624details
LFM2.5 1.2B Instruct1.2BF165.0 GBF1626details
HunyuanOCR1.1BF163.5 GBF1654details
PaddleOCR VL 1.6959MF162.5 GBF1663details
Qwen3.5 0.8B873MF162.4 GBF1679details
LightOnOCR 3 0.8B853MF162.4 GBF1679details
FastVLM 0.5B759MF162.2 GBF1679details
Qwen3 0.6B752MF162.7 GBF16100details
GOT OCR2 0716MF162.8 GBF1684details
Nemotron 3.5 Parse631MF161.9 GBF1692details
GOT OCR 2.0 hf561MF162.5 GBF16107details
SmolVLM2 500M Video Instruct507MF161.6 GBF16146details
SmolVLM 500M Instruct507MF161.6 GBF16146details
Qwen2.5 0.5B Instruct494MF161.7 GBF16121details
LFM2.5 VL 450M449MF161.9 GBF1684details
SmolLM2 360M Instruct362MF161.5 GBF16165details
SmolLM2 360M362MF161.5 GBF16165details
LFM2.5 350M354MF161.9 GBF1684details
LFM2 350M354MF161.4 GBF16169details

Just too big at 4-bit

These need up to 35% more memory than the Mac M4 32GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Qwen3.5 35B A3B21.5 GB0.1 GB
Yi 34B21.6 GB0.2 GB
Yi 34B 200K21.6 GB0.2 GB
Yi 34B Chat21.6 GB0.2 GB
Yi 1.5 34B Chat21.6 GB0.2 GB
Qwen AgentWorld 35B A3B21.6 GB0.2 GB
Maxwell V1 Flash22.1 GB0.6 GB
Qwen3.6 35B A3B22.1 GB0.7 GB
Nous Hermes 2 Yi 34B22.1 GB0.7 GB
Hermes 4.3 36B23.0 GB1.5 GB
Seed OSS 36B Instruct23.0 GB1.5 GB
flint23.6 GB2.1 GB
falcon 40b25.2 GB3.8 GB
Nous Hermes 2 Mixtral 8x7B DPO28.0 GB6.5 GB

Questions

What is the biggest model that runs on the Mac M4 32GB?

Of the models we track, the largest that fits at 8K context is Kimi Linear 48B A3B Instruct (49B) at Q2_K, needing about 17.3 GB.

How much of the Mac M4 32GB's memory can a model use?

About 21 GB of 32 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.