RunsOnMy ModelsGPUs & MacsBy VRAMAPI pricesVRAM calculatorNew

What AI models run on the Ryzen AI Max+ 395 128GB?

The Ryzen AI Max+ 395 128GB has 128 GB of unified memory (256 GB/s). About 109 GB is usable for a model. 235 of 285 open models we track fit at 8K context.

235models fit (8K ctx)
229still fit at 32K ctx
310Blargest that fits
256 GB/smemory bandwidth

Other 128 GB hardware: DGX Spark, Mac M3 Max 128GB, Mac M4 Max 128GB.

Models that fit, largest first

ModelParamsBest quant (8K)Needs32K contextMax tok/s (est.)
MiMo V2 Flash310B (16B act.)Q2_K106 GBno43details
DeepSeek V4 Flash Vision Exp305B (15B act.)Q2_K108 GBno45details
DeepSeek V4 Flash 0731304B (27B act.)Q2_K108 GBno25details
DeepSeek V4 Flash DSpark304B (27B act.)Q2_K108 GBno25details
Hy3299B (21B act.)Q2_K102 GBno34details
Hy3 preview299B (21B act.)Q2_K107 GBno33details
DeepSeek V4 Flash291B (14B act.)Q2_K103 GBQ2_K48details
ERNIE 5.5 Flash Base286B (5.7B act.)Q2_K101 GBQ2_K—details
Intern S1241B (28B act.)Q2_K86.2 GBQ2_K25details
DeepSeek V2 Chat236B (21B act.)Q2_K83.5 GBQ2_K32details
MiniMax M2.5229B (11B act.)Q2_K80.3 GBQ2_K64details
MiniMax M2.7229B (11B act.)Q3_K_M97.0 GBQ2_K52details
MiniMax M2229B (11B act.)Q2_K80.3 GBQ2_K64details
MiniMax M2.1229B (11B act.)Q2_K80.3 GBQ2_K64details
MiniMax M3.1 Flash220B (4.4B act.)Q2_K78.7 GBQ2_K—details
Step 3.7 Flash201BQ3_K_M91.6 GBQ2_K2.7details
Step 3.5 Flash199BQ3_K_M88.2 GBQ3_K_M2.8details
Qwen3.8 Flash Next180B (62B act.)Q3_K_M83.0 GBQ3_K_M8.5details
Ling 3.0 flash Sante127B (5.6B act.)Q5_K_M92.0 GBQ5_K_M60details
Ling 3.0 flash127B (5.6B act.)Q5_K_M85.5 GBQ5_K_M64details
Ornith 1.5 122B125B (13B act.)Q6_K96.6 GBQ6_K24details
gpt oss 120b117B (5.8B act.)F1661.9 GBF1680details
gpt oss safeguard 120b117B (5.8B act.)Q8_060.1 GBQ8_082details
GLM 4.5 Air110B (15B act.)Q6_K94.3 GBQ5_K_M19details
Llama 4 Scout 17B 16E Instruct109B (18B act.)Q6_K84.6 GBQ6_K17details
Llama 4 Scout 17B 16E109B (18B act.)Q6_K85.3 GBQ6_K17details
GLM 4.5V108B (14B act.)Q6_K84.4 GBQ6_K22details
GLM 4.6V108B (14B act.)Q6_K91.4 GBQ6_K20details
Hunyuan A13B Instruct80B (13B act.)Q8_081.2 GBQ8_019details
Kimi Dev 72B73BQ8_075.1 GBQ8_03.3details
Llama 3.3 70B Instruct71BQ8_073.0 GBQ8_03.4details
DeepSeek R1 Distill Llama 70B71BQ8_073.0 GBQ8_03.4details
Llama 3.1 70B71BQ8_073.0 GBQ8_03.4details
Kimi Linear 48B A3B Instruct49B (3.0B act.)F1692.5 GBF1643details
Nous Hermes 2 Mixtral 8x7B DPO47B (13B act.)F1688.5 GBF169.9details
falcon 40b42BF1679.4 GBF163.1details
flint40B (1.3B act.)F1675.1 GBF16102details
Maxwell V1 Flash37B (2.3B act.)F1669.7 GBF1656details
Hermes 4.3 36B36BF1670.0 GBF163.5details
Seed OSS 36B Instruct36BF1670.0 GBF163.5details
Qwen3.6 35B A3B36B (3.0B act.)F1667.2 GBF1643details
Ornith 1.5 35B A3B36B (3.0B act.)F1667.2 GBF1643details
Qwen3.5 35B A3B36B (3.0B act.)F1665.6 GBF1644details
Axion 1.0 35B35B (3.9B act.)F1665.6 GBF1633details
Qwen AgentWorld 35B A3B35B (3.0B act.)F1665.6 GBF1643details
Yi 34B 200K34BF1666.5 GBF163.7details
Yi 34B34BF1666.5 GBF163.7details
Yi 34B Chat34BF1666.5 GBF163.7details
Yi 1.5 34B Chat34BF1666.5 GBF163.7details
Nous Hermes 2 Yi 34B34BF1666.5 GBF163.7details
DeepSeek R1 Distill Qwen 32B33BF1663.7 GBF163.9details
Qwen3 32B33BF1663.7 GBF163.9details
GLM Z1 32B 041433BF1661.8 GBF163.9details
GLM 4 32B 041433BF1661.8 GBF163.9details
granite 4.0 h small32B (8.8B act.)F1660.7 GBF1615details
gemma 4 31B it31BF1660.1 GBF164.2details
GLM 4.7 Flash31B (4.6B act.)F1656.9 GBF1629details
Qwen3 Coder 30B A3B Instruct31B (3.0B act.)F1658.3 GBF1643details
North Mini Code 1.030B (3.3B act.)F1658.2 GBF1639details
Hy Memory 30B A3B30B (3.0B act.)F1657.4 GBF1643details
Muse Glimmer 30B30BF1652.9 GBF164.6details
granite 4.2 30b29BF1657.1 GBF164.4details
Qwen3.8 27B28BF1652.3 GBF164.7details
Qwen3.6 27B28BF1651.5 GBF164.8details
Qwen3.5 27B28BF1651.5 GBF164.8details
gemma 3 27b it27BF1652.2 GBF164.7details
Arete 27B27BF1651.5 GBF164.8details
MiniMax H3 Image PE27BF1652.3 GBF164.7details
Trinity Mini26B (2.0B act.)F1649.7 GBF1665details
GLiDE 27B26BF1650.0 GBF164.9details
diffusiongemma 26B A4B it26B (4.0B act.)F1648.5 GBF1633details
gemma 4 26B A4B it26B (4.0B act.)F1648.4 GBF1633details
LFM2.5 24B A2B24B (2.0B act.)F1690.1 GBF1632details
ERNIE 4.5 21B A3B PT22BF1641.7 GBF165.9details
ERNIE 4.5 21B A3B Thinking22BF1641.7 GBF165.9details
gpt oss safeguard 20b22B (4.8B act.)F1613.8 GBF1683details
gpt oss 20b21B (4.2B act.)F1613.8 GBF1693details
Groot 19B A3B 26102319B (3.0B act.)F1636.4 GBF1644details
Kimi VL A3B Thinking 250616B (3.0B act.)F1660.4 GBF1622details
Kimi VL A3B Instruct16B (3.0B act.)F1631.5 GBF1643details
Kimi VL A3B Thinking16B (3.0B act.)F1631.5 GBF1643details
deepseek moe 16b base16B (2.8B act.)F1632.8 GBF1645details
LLaDA2.1 mini16B (1.4B act.)F1631.3 GBF1689details
Moonlight 16B A3B Instruct16B (3.0B act.)F1630.7 GBF1643details
DeepSeek Coder V2 Lite Instruct16B (2.7B act.)F1630.1 GBF1648details
DeepSeek V2 Lite16B (2.7B act.)F1630.1 GBF1648details
DeepSeek V2 Lite Chat16B (2.7B act.)F1630.1 GBF1648details
Qwen2.5 Coder 14B Instruct15BF1629.7 GBF168.7details
DeepSeek R1 Distill Qwen 14B15BF1629.7 GBF168.7details
Qwen3 14B15BF1629.4 GBF168.7details
phi 415BF1629.5 GBF168.7details
gemma 3 12b it12BF1623.6 GBF1611details
Llama Guard 4 12B12BF1624.6 GBF1611details
Llama 3.2 11B Vision Instruct11BF1620.1 GBF1613details
Llama 3.2 11B Vision11BF1621.7 GBF1612details
GLM 4.6V Flash10BF1618.5 GBF1614details
GLM 4.1V 9B Thinking10BF1618.5 GBF1614details
Step3 VL 10B10BF1632.3 GBF167.8details
SenseNova Loop 9B Preview9.7BF1617.8 GBF1614details
Qwen3.5 9B9.7BF1617.8 GBF1614details
Ornith 1.5 9B9.7BF1617.8 GBF1614details
Qwen3.5 9B Base9.7BF1617.8 GBF1614details
MiMo V2.6 Distill Qwen 9B9.4BF1617.8 GBF1614details
glm 4 9b9.4BF1623.2 GBF1614details
internlm3 8b instruct8.8BF1617.4 GBF1615details
granite 4.2 8b8.8BF1618.2 GBF1615details
granite 4.1 8b8.8BF1618.2 GBF1615details
Qwen3 VL 8B Instruct8.8BF1617.1 GBF1616details
Molmo2 8B8.7BF1617.1 GBF1616details
LFM2.5 8B A1B8.5B (1.0B act.)F1632.3 GBF1664details
Qwen2.5 VL 7B Instruct8.3BF1615.3 GBF1617details
UI TARS 1.5 7B8.3BF1615.3 GBF1617details
olmOCR 2 7B 10258.3BF1615.3 GBF1617details
olmOCR 7B 0225 preview8.3BF1616.6 GBF1615details
UI TARS 7B DPO8.3BF1615.3 GBF1617details
Qwen3 8B8.2BF1617.1 GBF1616details
DeepSeek R1 0528 Qwen3 8B8.2BF1617.1 GBF1616details
granite 3.3 8b instruct8.2BF1617.0 GBF1616details
granite 3.1 8b instruct8.2BF1617.0 GBF1616details
granite 3.2 8b instruct8.2BF1617.0 GBF1616details
Apertus 8B Instruct 25098.1BF1616.7 GBF1616details
Llama 3.1 8B Instruct8.0BF1616.6 GBF1616details
Llama 3.1 8B8.0BF1616.6 GBF1616details
Hermes 3 Llama 3.1 8B8.0BF1616.6 GBF1616details
DeepSeek R1 Distill Llama 8B8.0BF1631.6 GBF168.0details
Meta Llama 3 8B Instruct8.0BF1616.6 GBF1616details
Meta Llama Guard 2 8B8.0BF1616.6 GBF1616details
Meta Llama 3.1 8B Instruct8.0BF1616.6 GBF1616details
Ling 3.0 tiny7.9B (1.4B act.)F1615.6 GBF1693details
MiMo 7B RL7.8BF1616.4 GBF1616details
internlm2 5 7b chat7.7BF1616.0 GBF1617details
internlm2 chat 7b7.7BF1616.0 GBF1617details
Qwen2.5 7B Instruct7.6BF1615.3 GBF1617details
Qwen2.5 Coder 7B Instruct7.6BF1615.3 GBF1617details
DeepSeek R1 Distill Qwen 7B7.6BF1615.3 GBF1617details
Falcon H1R 7B7.6BF1615.1 GBF1617details
Olmo 3 7B Instruct7.3BF1616.7 GBF1618details
Olmo 3 1025 7B7.3BF1616.7 GBF1618details
Olmo 3 7B Think7.3BF1616.7 GBF1618details
Mistral 7B v0.17.2BF1614.9 GBF1618details
Mistral 7B Instruct v0.27.2BF1614.9 GBF1618details
Mistral 7B Instruct v0.17.2BF1614.9 GBF1618details
falcon 7b7.2BF1618.4 GBF1618details
falcon 7b instruct7.2BF1618.4 GBF1618details
granite 4.0 h tiny6.9B (1.5B act.)F1613.6 GBF1687details
OLMoE 1B 7B 09246.9B (1.3B act.)F1614.4 GBF16100details
deepseek coder 7b instruct v1.56.9BF1617.2 GBF1619details
deepseek coder 6.7b instruct6.7BF1617.0 GBF1619details
deepseek coder 6.7b base6.7BF1617.0 GBF1619details
Llama 2 7b hf6.7BF1617.0 GBF1619details
Yi 6B 200K6.1BF1612.3 GBF1621details
Yi 6B6.1BF1612.3 GBF1621details
Molmo2 4B4.9BF1610.0 GBF1629details
WeVisOne VL 4B4.8BF1610.0 GBF1629details
Mage VL Cowork4.7BF1610.0 GBF1629details
Qianfan OCR4.7BF1610.1 GBF1629details
Qwen3.5 4B4.7BF169.0 GBF1630details
S5 nano4.7BF169.2 GBF1630details
LightOnOCR 3 4B4.5BF169.0 GBF1630details
SingVL S1 4B4.5BF169.0 GBF1630details
Intern Decision 4B4.5BF169.0 GBF1630details
Qwen3 VL 4B Instruct4.4BF169.3 GBF1632details
gemma 3 4b it4.3BF168.4 GBF1633details
medgemma 1.5 4b it4.3BF168.4 GBF1633details
medgemma 4b it4.3BF168.4 GBF1633details
gemma 3 4b pt4.3BF169.2 GBF1630details
Phi 3.5 vision instruct4.1BF1611.2 GBF1631details
Qwen3 4B4.0BF169.3 GBF1632details
Qwen3 4B Instruct 25074.0BF169.3 GBF1632details
granite vision 4.1 4b4.0BF167.6 GBF1638details
Phi 4 mini instruct3.8BF168.9 GBF1633details
Phi 3 mini 4k instruct3.8BF1610.6 GBF1633details
Phi 3.5 mini instruct3.8BF1610.6 GBF1633details
Phi 3 mini 128k instruct3.8BF1610.6 GBF1633details
Qwen2.5 VL 3B Instruct3.8BF166.7 GBF1641details
granite 4.2 3b3.7BF168.0 GBF1635details
granite 4.1 3b llmon preview3.7BF168.0 GBF1635details
Raon 1 nano3.5BF167.0 GBF1641details
granite 4.1 3b3.4BF167.6 GBF1638details
granite 4.0 micro3.4BF167.6 GBF1638details
DeepSeek OCR 23.4B (1.2B act.)F166.6 GBF16124details
deepseek vl2 tiny3.4B (1.2B act.)F167.4 GBF16109details
Unlimited OCR 23.3B (1.1B act.)F166.6 GBF16127details
Unlimited OCR3.3B (1.1B act.)F166.6 GBF16127details
DeepSeek OCR3.3B (1.1B act.)F166.6 GBF16127details
granite 3.1 3b a800m instruct3.3B (883M act.)F167.1 GBF16145details
Llama 3.2 3B Instruct3.2BF1613.5 GBF1620details
Llama 3.2 3B3.2BF167.5 GBF1640details
Hermes 3 Llama 3.2 3B3.2BF167.5 GBF1640details
d1 3B3.1BF1610.8 GBF1624details
LFM2.5 VL 3B3.1BF1610.8 GBF1624details
Qwen2.5 3B Instruct3.1BF166.7 GBF1641details
SmolLM3 3B3.1BF166.9 GBF1642details
SmolLM3 3B Base3.1BF166.9 GBF1642details
phi 22.8BF168.2 GBF1646details
LFM2.5 2.6B2.7BF1610.8 GBF1624details
granite 3.3 2b instruct2.5BF165.8 GBF1651details
MiniCPM5 2B2.5BF165.7 GBF1651details
WeVisOne VL 2B2.4BF165.4 GBF1663details
Qwen3.5 2B2.3BF164.5 GBF1668details
SmolVLM Instruct2.2BF165.4 GBF1671details
Qwen3 1.7B2.0BF164.8 GBF1674details
Qwen3.8 27B DSpine b162.0BF164.8 GBF1663details
d1 2B1.9BF167.2 GBF1637details
LFM3 VL 2B1.9BF167.2 GBF1637details
DeepSeek R1 Distill Qwen 1.5B1.8BF164.2 GBF1672details
LFM3 1.7B1.7BF167.2 GBF1637details
LFM3 1.7B Base1.7BF164.0 GBF1674details
SmolLM2 1.7B Instruct1.7BF165.2 GBF1675details
SmolLM2 1.7B1.7BF165.2 GBF1675details
SmolLM 1.7B1.7BF165.2 GBF1675details
granite 4.0 1b1.6BF164.3 GBF1678details
Qwen2.5 1.5B Instruct1.5BF163.8 GBF1683details
OLMo 2 0425 1B1.5BF164.4 GBF1686details
deepseek coder 1.3b instruct1.3BF164.5 GBF1695details
Llama 3.2 1B1.2BF163.2 GBF16104details
Llama 3.2 1B Instruct1.2BF165.5 GBF1652details
LFM2.5 1.2B Instruct1.2BF165.0 GBF1655details
HunyuanOCR1.1BF163.5 GBF16114details
PaddleOCR VL 1.6959MF162.5 GBF16134details
Qwen3.5 0.8B873MF162.4 GBF16169details
LightOnOCR 3 0.8B853MF162.4 GBF16169details
FastVLM 0.5B759MF162.2 GBF16169details
Qwen3 0.6B752MF162.7 GBF16214details
GOT OCR2 0716MF162.8 GBF16179details
Nemotron 3.5 Parse631MF161.9 GBF16196details
GOT OCR 2.0 hf561MF162.5 GBF16228details
SmolVLM2 500M Video Instruct507MF161.6 GBF16312details
SmolVLM 500M Instruct507MF161.6 GBF16312details
Qwen2.5 0.5B Instruct494MF161.7 GBF16259details
LFM2.5 VL 450M449MF161.9 GBF16180details
SmolLM2 360M Instruct362MF161.5 GBF16353details
SmolLM2 360M362MF161.5 GBF16353details
LFM2.5 350M354MF161.9 GBF16180details
LFM2 350M354MF161.4 GBF16361details

Just too big at 4-bit

These need up to 35% more memory than the Ryzen AI Max+ 395 128GB has at Q4_K_M and 8K context. A smaller quant or shorter context may work.

ModelQ4_K_M needsShort by
Step 3.5 Flash115.9 GB7.1 GB
Step 3.7 Flash118.8 GB10.0 GB
MiniMax M3.1 Flash127.2 GB18.4 GB
MiniMax M2131.6 GB22.8 GB
MiniMax M2.5131.6 GB22.8 GB
MiniMax M2.1131.6 GB22.8 GB
MiniMax M2.7133.3 GB24.5 GB
DeepSeek V2 Chat135.3 GB26.5 GB
Intern S1139.2 GB30.4 GB

Questions

What is the biggest model that runs on the Ryzen AI Max+ 395 128GB?

Of the models we track, the largest that fits at 8K context is MiMo V2 Flash (310B) at Q2_K, needing about 106 GB.

How much of the Ryzen AI Max+ 395 128GB's memory can a model use?

About 109 GB of 128 GB (we assume ~85% of shared memory can be given to the GPU).

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