RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

What AI models run on the Mac M3 Ultra 256GB?

The Mac M3 Ultra 256GB has 256 GB of unified memory (819 GB/s). About 192 GB is usable for a model. 248 of 285 open models we track fit at 8K context.

248models fit (8K ctx)
247still fit at 32K ctx
456Blargest that fits
819 GB/smemory bandwidth

Apple Silicon uses unified memory; see the Mac guide for default macOS limits across machines.

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
MiniMax M1 40k456B (48B act.)Q2_K163 GBQ2_K45details
MiniMax Text 01456B (48B act.)Q2_K163 GBQ2_K45details
MiniMax M3427B (20B act.)Q2_K151 GBQ2_K110details
Llama 4 Maverick 17B 128E Instruct402B (18B act.)Q2_K138 GBQ2_K125details
GLM 4.5358B (36B act.)Q3_K_M163 GBQ3_K_M48details
GLM 4.7358B (36B act.)Q3_K_M163 GBQ3_K_M48details
GLM 4.6357B (34B act.)Q3_K_M162 GBQ3_K_M50details
GLM 5.3 Flash321B (18B act.)Q3_K_M147 GBQ3_K_M91details
Nex N2.6321B (18B act.)Q3_K_M145 GBQ3_K_M93details
step3321BQ2_K143 GBno6.8details
MiMo V2.6 Flash RL311B (6.2B act.)Q3_K_M143 GBQ3_K_M—details
MiMo V2.5311B (17B act.)Q3_K_M132 GBQ3_K_M104details
MiMo V2.6 Flash MOPD311B (6.2B act.)Q3_K_M143 GBQ3_K_M—details
MiMo V2 Flash310B (16B act.)Q3_K_M139 GBQ3_K_M105details
DeepSeek V4 Flash Vision Exp305B (15B act.)Q3_K_M140 GBQ3_K_M110details
DeepSeek V4 Flash 0731304B (27B act.)Q3_K_M140 GBQ3_K_M61details
DeepSeek V4 Flash DSpark304B (27B act.)Q3_K_M140 GBQ3_K_M61details
Hy3299B (21B act.)Q4_K_M173 GBQ3_K_M64details
Hy3 preview299B (21B act.)Q3_K_M139 GBQ3_K_M80details
DeepSeek V4 Flash291B (14B act.)Q4_K_M167 GBQ4_K_M95details
ERNIE 5.5 Flash Base286B (5.7B act.)Q4_K_M164 GBQ4_K_M—details
Intern S1241B (28B act.)Q5_K_M162 GBQ5_K_M41details
DeepSeek V2 Chat236B (21B act.)Q5_K_M158 GBQ5_K_M54details
MiniMax M2.5229B (11B act.)Q5_K_M154 GBQ5_K_M105details
MiniMax M2.7229B (11B act.)Q5_K_M160 GBQ5_K_M101details
MiniMax M2229B (11B act.)Q5_K_M154 GBQ5_K_M105details
MiniMax M2.1229B (11B act.)Q5_K_M154 GBQ5_K_M105details
MiniMax M3.1 Flash220B (4.4B act.)Q6_K170 GBQ5_K_M—details
Step 3.7 Flash201BQ6_K158 GBQ6_K5.0details
Step 3.5 Flash199BQ6_K155 GBQ6_K5.1details
Qwen3.8 Flash Next180B (62B act.)Q6_K139 GBQ6_K16details
Ling 3.0 flash Sante127B (5.6B act.)Q8_0127 GBQ8_0137details
Ling 3.0 flash127B (5.6B act.)Q8_0127 GBQ8_0137details
Ornith 1.5 122B125B (13B act.)Q8_0125 GBQ8_061details
gpt oss 120b117B (5.8B act.)F1661.9 GBF16255details
gpt oss safeguard 120b117B (5.8B act.)Q8_060.1 GBQ8_0262details
GLM 4.5 Air110B (15B act.)Q8_0112 GBQ8_051details
Llama 4 Scout 17B 16E Instruct109B (18B act.)Q8_0109 GBQ8_043details
Llama 4 Scout 17B 16E109B (18B act.)Q8_0110 GBQ8_043details
GLM 4.5V108B (14B act.)Q8_0108 GBQ8_054details
GLM 4.6V108B (14B act.)Q8_0108 GBQ8_054details
Hunyuan A13B Instruct80B (13B act.)F16151 GBF1631details
Kimi Dev 72B73BF16139 GBF165.6details
Llama 3.3 70B Instruct71BF16135 GBF165.8details
DeepSeek R1 Distill Llama 70B71BF16135 GBF165.8details
Llama 3.1 70B71BF16135 GBF165.8details
Kimi Linear 48B A3B Instruct49B (3.0B act.)F1692.5 GBF16136details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)F1688.5 GBF1632details
falcon 40b42BF1679.4 GBF169.8details
flint40B (1.3B act.)F1675.1 GBF16328details
Maxwell V1 Flash37B (2.3B act.)F1669.7 GBF16180details
Hermes 4.3 36B36BF1670.0 GBF1611details
Seed OSS 36B Instruct36BF1670.0 GBF1611details
Qwen3.6 35B A3B36B (3.0B act.)F1667.2 GBF16138details
Ornith 1.5 35B A3B36B (3.0B act.)F1667.2 GBF16138details
Qwen3.5 35B A3B36B (3.0B act.)F1665.6 GBF16141details
Axion 1.0 35B35B (3.9B act.)F1665.6 GBF16106details
Qwen AgentWorld 35B A3B35B (3.0B act.)F1665.6 GBF16136details
Yi 34B 200K34BF1666.5 GBF1612details
Yi 34B34BF1666.5 GBF1612details
Yi 34B Chat34BF1666.5 GBF1612details
Yi 1.5 34B Chat34BF1666.5 GBF1612details
Nous Hermes 2 Yi 34B34BF1666.5 GBF1612details
DeepSeek R1 Distill Qwen 32B33BF1663.7 GBF1612details
Qwen3 32B33BF1663.7 GBF1612details
GLM Z1 32B 041433BF1661.8 GBF1613details
GLM 4 32B 041433BF1661.8 GBF1613details
granite 4.0 h small32B (8.8B act.)F1660.7 GBF1646details
gemma 4 31B it31BF1660.1 GBF1613details
GLM 4.7 Flash31B (4.6B act.)F1656.9 GBF1693details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)F1658.3 GBF16136details
North Mini Code 1.030B (3.3B act.)F1658.2 GBF16124details
Hy Memory 30B A3B30B (3.0B act.)F1657.4 GBF16137details
Muse Glimmer 30B30BF1652.9 GBF1615details
granite 4.2 30b29BF1657.1 GBF1614details
Qwen3.8 27B28BF1652.3 GBF1615details
Qwen3.6 27B28BF1651.5 GBF1615details
Qwen3.5 27B28BF1651.5 GBF1615details
gemma 3 27b it27BF1652.2 GBF1615details
Arete 27B27BF1651.5 GBF1615details
MiniMax H3 Image PE27BF1652.3 GBF1615details
Trinity Mini26B (2.0B act.)F1649.7 GBF16208details
GLiDE 27B26BF1650.0 GBF1616details
diffusiongemma 26B A4B it26B (4.0B act.)F1648.5 GBF16105details
gemma 4 26B A4B it26B (4.0B act.)F1648.4 GBF16105details
LFM2.5 24B A2B24B (2.0B act.)F1690.1 GBF16102details
ERNIE 4.5 21B A3B PT22BF1641.7 GBF1619details
ERNIE 4.5 21B A3B Thinking22BF1641.7 GBF1619details
gpt oss safeguard 20b22B (4.8B act.)F1613.8 GBF16267details
gpt oss 20b21B (4.2B act.)F1613.8 GBF16296details
Groot 19B A3B 26102319B (3.0B act.)F1636.4 GBF16140details
Kimi VL A3B Thinking 250616B (3.0B act.)F1660.4 GBF1670details
Kimi VL A3B Instruct16B (3.0B act.)F1631.5 GBF16137details
Kimi VL A3B Thinking16B (3.0B act.)F1631.5 GBF16137details
deepseek moe 16b base16B (2.8B act.)F1632.8 GBF16145details
LLaDA2.1 mini16B (1.4B act.)F1631.3 GBF16286details
Moonlight 16B A3B Instruct16B (3.0B act.)F1630.7 GBF16137details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)F1630.1 GBF16154details
DeepSeek V2 Lite16B (2.7B act.)F1630.1 GBF16154details
DeepSeek V2 Lite Chat16B (2.7B act.)F1630.1 GBF16154details
Qwen2.5 Coder 14B Instruct15BF1629.7 GBF1628details
DeepSeek R1 Distill Qwen 14B15BF1629.7 GBF1628details
Qwen3 14B15BF1629.4 GBF1628details
phi 415BF1629.5 GBF1628details
gemma 3 12b it12BF1623.6 GBF1635details
Llama Guard 4 12B12BF1624.6 GBF1634details
Llama 3.2 11B Vision Instruct11BF1620.1 GBF1642details
Llama 3.2 11B Vision11BF1621.7 GBF1638details
GLM 4.6V Flash10BF1618.5 GBF1644details
GLM 4.1V 9B Thinking10BF1618.5 GBF1644details
Step3 VL 10B10BF1632.3 GBF1625details
SenseNova Loop 9B Preview9.7BF1617.8 GBF1646details
Qwen3.5 9B9.7BF1617.8 GBF1646details
Ornith 1.5 9B9.7BF1617.8 GBF1646details
Qwen3.5 9B Base9.7BF1617.8 GBF1646details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF1646details
glm 4 9b9.4BF1623.2 GBF1644details
internlm3 8b instruct8.8BF1617.4 GBF1647details
granite 4.2 8b8.8BF1618.2 GBF1647details
granite 4.1 8b8.8BF1618.2 GBF1647details
Qwen3 VL 8B Instruct8.8BF1617.1 GBF1650details
Molmo2 8B8.7BF1617.1 GBF1650details
LFM2.5 8B A1B8.5B (1.0B act.)F1632.3 GBF16205details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF1654details
UI TARS 1.5 7B8.3BF1615.3 GBF1654details
olmOCR 2 7B 10258.3BF1615.3 GBF1654details
olmOCR 7B 0225 preview8.3BF1616.6 GBF1649details
UI TARS 7B DPO8.3BF1615.3 GBF1654details
Qwen3 8B8.2BF1617.1 GBF1650details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBF1650details
granite 3.3 8b instruct8.2BF1617.0 GBF1650details
granite 3.1 8b instruct8.2BF1617.0 GBF1650details
granite 3.2 8b instruct8.2BF1617.0 GBF1650details
Apertus 8B Instruct 25098.1BF1616.7 GBF1651details
Llama 3.1 8B Instruct8.0BF1616.6 GBF1651details
Llama 3.1 8B8.0BF1616.6 GBF1651details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBF1651details
DeepSeek R1 Distill Llama 8B8.0BF1631.6 GBF1625details
Meta Llama 3 8B Instruct8.0BF1616.6 GBF1651details
Meta Llama Guard 2 8B8.0BF1616.6 GBF1651details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBF1651details
Ling 3.0 tiny7.9B (1.4B act.)F1615.6 GBF16296details
MiMo 7B RL7.8BF1616.4 GBF1652details
internlm2 5 7b chat7.7BF1616.0 GBF1653details
internlm2 chat 7b7.7BF1616.0 GBF1653details
Qwen2.5 7B Instruct7.6BF1615.3 GBF1654details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF1654details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF1654details
Falcon H1R 7B7.6BF1615.1 GBF1654details
Olmo 3 7B Instruct7.3BF1616.7 GBF1656details
Olmo 3 1025 7B7.3BF1616.7 GBF1656details
Olmo 3 7B Think7.3BF1616.7 GBF1656details
Mistral 7B v0.17.2BF1614.9 GBF1657details
Mistral 7B Instruct v0.27.2BF1614.9 GBF1657details
Mistral 7B Instruct v0.17.2BF1614.9 GBF1657details
falcon 7b7.2BF1618.4 GBF1657details
falcon 7b instruct7.2BF1618.4 GBF1657details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBF16279details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF16319details
deepseek coder 7b instruct v1.56.9BF1617.2 GBF1659details
deepseek coder 6.7b instruct6.7BF1617.0 GBF1661details
deepseek coder 6.7b base6.7BF1617.0 GBF1661details
Llama 2 7b hf6.7BF1617.0 GBF1661details
Yi 6B 200K6.1BF1612.3 GBF1668details
Yi 6B6.1BF1612.3 GBF1668details
Molmo2 4B4.9BF1610.0 GBF1693details
WeVisOne VL 4B4.8BF1610.0 GBF1693details
Mage VL Cowork4.7BF1610.0 GBF1693details
Qianfan OCR4.7BF1610.1 GBF1693details
Qwen3.5 4B4.7BF169.0 GBF1697details
S5 nano4.7BF169.2 GBF1695details
LightOnOCR 3 4B4.5BF169.0 GBF1697details
SingVL S1 4B4.5BF169.0 GBF1697details
Intern Decision 4B4.5BF169.0 GBF1697details
Qwen3 VL 4B Instruct4.4BF169.3 GBF16102details
gemma 3 4b it4.3BF168.4 GBF16105details
medgemma 1.5 4b it4.3BF168.4 GBF16105details
medgemma 4b it4.3BF168.4 GBF16105details
gemma 3 4b pt4.3BF169.2 GBF1695details
Phi 3.5 vision instruct4.1BF1611.2 GBF1699details
Qwen3 4B4.0BF169.3 GBF16102details
Qwen3 4B Instruct 25074.0BF169.3 GBF16102details
granite vision 4.1 4b4.0BF167.6 GBF16120details
Phi 4 mini instruct3.8BF168.9 GBF16107details
Phi 3 mini 4k instruct3.8BF1610.6 GBF16107details
Phi 3.5 mini instruct3.8BF1610.6 GBF16107details
Phi 3 mini 128k instruct3.8BF1610.6 GBF16107details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF16133details
granite 4.2 3b3.7BF168.0 GBF16112details
granite 4.1 3b llmon preview3.7BF168.0 GBF16112details
Raon 1 nano3.5BF167.0 GBF16131details
granite 4.1 3b3.4BF167.6 GBF16120details
granite 4.0 micro3.4BF167.6 GBF16120details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16396details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF16348details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16405details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16407details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16407details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF16463details
Llama 3.2 3B Instruct3.2BF1613.5 GBF1664details
Llama 3.2 3B3.2BF167.5 GBF16127details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF16127details
d1 3B3.1BF1610.8 GBF1676details
LFM2.5 VL 3B3.1BF1610.8 GBF1676details
Qwen2.5 3B Instruct3.1BF166.7 GBF16133details
SmolLM3 3B3.1BF166.9 GBF16133details
SmolLM3 3B Base3.1BF166.9 GBF16133details
phi 22.8BF168.2 GBF16147details
LFM2.5 2.6B2.7BF1610.8 GBF1676details
granite 3.3 2b instruct2.5BF165.8 GBF16162details
MiniCPM5 2B2.5BF165.7 GBF16163details
WeVisOne VL 2B2.4BF165.4 GBF16201details
Qwen3.5 2B2.3BF164.5 GBF16217details
SmolVLM Instruct2.2BF165.4 GBF16226details
Qwen3 1.7B2.0BF164.8 GBF16238details
Qwen3.8 27B DSpine b162.0BF164.8 GBF16202details
d1 2B1.9BF167.2 GBF16118details
LFM3 VL 2B1.9BF167.2 GBF16118details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF16230details
LFM3 1.7B1.7BF167.2 GBF16118details
LFM3 1.7B Base1.7BF164.0 GBF16237details
SmolLM2 1.7B Instruct1.7BF165.2 GBF16239details
SmolLM2 1.7B1.7BF165.2 GBF16239details
SmolLM 1.7B1.7BF165.2 GBF16239details
granite 4.0 1b1.6BF164.3 GBF16251details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF16265details
OLMo 2 0425 1B1.5BF164.4 GBF16275details
deepseek coder 1.3b instruct1.3BF164.5 GBF16304details
Llama 3.2 1B1.2BF163.2 GBF16331details
Llama 3.2 1B Instruct1.2BF165.5 GBF16165details
LFM2.5 1.2B Instruct1.2BF165.0 GBF16175details
HunyuanOCR1.1BF163.5 GBF16366details
PaddleOCR VL 1.6959MF162.5 GBF16427details
Qwen3.5 0.8B873MF162.4 GBF16540details
LightOnOCR 3 0.8B853MF162.4 GBF16540details
FastVLM 0.5B759MF162.2 GBF16540details
Qwen3 0.6B752MF162.7 GBF16684details
GOT OCR2 0716MF162.8 GBF16572details
Nemotron 3.5 Parse631MF161.9 GBF16627details
GOT OCR 2.0 hf561MF162.5 GBF16731details
SmolVLM2 500M Video Instruct507MF161.6 GBF16998details
SmolVLM 500M Instruct507MF161.6 GBF16998details
Qwen2.5 0.5B Instruct494MF161.7 GBF16829details
LFM2.5 VL 450M449MF161.9 GBF16576details
SmolLM2 360M Instruct362MF161.5 GBF161129details
SmolLM2 360M362MF161.5 GBF161129details
LFM2.5 350M354MF161.9 GBF16576details
LFM2 350M354MF161.4 GBF161155details

Just too big at 4-bit

These need up to 35% more memory than the Mac M3 Ultra 256GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
GLM 4.6204.3 GB12.3 GB
GLM 4.7205.2 GB13.2 GB
GLM 4.5205.2 GB13.2 GB
step3213.9 GB21.9 GB
Llama 4 Maverick 17B 128E Instruct228.4 GB36.4 GB
MiniMax M3248.5 GB56.5 GB

Questions

What is the biggest model that runs on the Mac M3 Ultra 256GB?

Of the models we track, the largest that fits at 8K context is MiniMax M1 40k (456B) at Q2_K, needing about 163 GB.

How much of the Mac M3 Ultra 256GB's memory can a model use?

About 192 GB of 256 GB. macOS lets the GPU use about 75% 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.