AMD · unified x86

AMD Ryzen AI Max+ 395 (Radeon 8060S)

AMD Ryzen AI Max+ 395 (Radeon 8060S) has 64 GB of VRAM at 256 GB/s — about 44.64 GiB usable after driver and compositor overhead. 2031 of 2118 indexed models fit at 32K context with f16 KV. Note only 48 GB of its 64 GB is allocatable to the GPU.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
64 GB
LPDDR5X-8000
Bandwidth
256 GB/s
256-bit bus
Tensor FP16
—
dense
TDP
120 W
Context4K8K16K32K64K128K
KV cachef16q8_0q4_0quantizing the KV cache is a ~2× lever on the dominant term at long context
text 1743vision language 184image 2audio tts 21video 16audio asr 39embedding 26

What fits at 32K context

largest quantization that fits, per model · 2031 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
command-a-plus-05-2026-bf16MoEIQ1_S219B45.87 GiB1.42 GiB48.00 GiB0.00 GiB12±37%
Llama-4-Scout-17B-16E-Instruct-abliterated-v2MoEKV unresolvedI1-IQ3_XS109B41.25 GiB6.00 GiB47.98 GiB0.02 GiB7±37%
Ornith-1.0-35B-AEON-Ultimate-Uncensored-BF16MoEQ5_K_M35.1B46.64 GiB0.63 GiB47.97 GiB0.03 GiB16±37%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEIQ1_M124B44.51 GiB2.75 GiB47.96 GiB0.04 GiB11±37%
Qwen3.5-88BMoEI1-Q4_087.7B46.41 GiB0.75 GiB47.89 GiB0.11 GiB14±37%
Mistral-Small-4-119B-2603MoEIQ3_XXS119B46.44 GiB0.70 GiB47.88 GiB0.12 GiB16±37%
Kimi-Dev-72BIQ4_XS72.7B37.02 GiB10.00 GiB47.85 GiB0.15 GiB3±25%
Qwen2.5-VL-72B-InstructIQ4_XS73.4B37.02 GiB10.00 GiB47.85 GiB0.15 GiB3±25%
GLM-4.5-AirMoEUD-IQ2_M110B41.34 GiB5.75 GiB47.82 GiB0.18 GiB8±37%
Rombo-LLM-V3.0-Qwen-72bI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Qwen2.5-72B-Instruct-abliteratedI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Qwen2.5-72B-Instruct-abliterated-v2I1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
HuatuoGPT-o1-72BIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
MiroThinker-v1.0-72BI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
EVA-Qwen2.5-72B-v0.2IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Qwen2.5-Math-72B-InstructIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Qwen2.5-72B-InstructIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Malaysian-Qwen2.5-72B-InstructI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Qwen2.5-72BI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
magnum-v4-72bI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
KAT-Dev-72B-ExpIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Homer-v1.0-Qwen2.5-72BIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Tower-Plus-72B-ultra-uncensored-hereticI1-IQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
Chronos-Platinum-72BIQ4_XS72.7B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
UI-TARS-72B-DPOIQ4_XS73.4B36.98 GiB10.00 GiB47.81 GiB0.19 GiB3±25%
HunyuanImage-2.1Q5_017.5B47.04 GiB0.00 GiB47.79 GiB0.21 GiB3±25%
Apertus-70B-Instruct-2509Q3_K_L70.6B36.87 GiB10.00 GiB47.75 GiB0.25 GiB3±25%
Llama-4-Scout-17B-16E-InstructMoEKV unresolvedQ2_K_L109B40.97 GiB6.00 GiB47.70 GiB0.30 GiB7±37%
Qwen3-Coder-Next-REAMMoEI1-Q6_K60.3B46.22 GiB0.75 GiB47.66 GiB0.34 GiB17±37%
Chuluun-Qwen2.5-72B-v0.01Q3_K_L72.7B36.79 GiB10.00 GiB47.62 GiB0.38 GiB3±25%
Qwen3-72B-SynthesisQ3_K_L72.7B36.79 GiB10.00 GiB47.62 GiB0.38 GiB3±25%
Behemoth-X-123B-v2IQ2_S123B35.75 GiB11.00 GiB47.60 GiB0.40 GiB3±25%
Delphi-25B-SimpleRL-MathI1-Q4_K_S25.0B13.35 GiB33.47 GiB47.59 GiB0.41 GiB3±25%
Huihui-GLM-4.7-Flash-abliterated-57BMoEI1-Q6_K57.3B42.64 GiB4.18 GiB47.57 GiB0.43 GiB9±37%
Noromaid-20b-v0.1.1I1-IQ3_S20.0B8.06 GiB38.75 GiB47.56 GiB0.44 GiB3±25%
Nethena-20BQ3_K_S20.0B8.06 GiB38.75 GiB47.56 GiB0.44 GiB3±25%
Laguna-S-2.1MoEUD-IQ3_S118B45.10 GiB1.64 GiB47.47 GiB0.53 GiB13±37%
Hunyuan-A13B-InstructMoEIQ4_NL80.4B42.77 GiB4.00 GiB47.46 GiB0.54 GiB3±25%
Qwen3-Coder-NextMoEQ4_K_S79.7B43.71 GiB3.00 GiB47.40 GiB0.60 GiB12±37%
Qwen3-Next-80B-A3B-ThinkingMoEQ4_K_S81.3B43.71 GiB3.00 GiB47.40 GiB0.60 GiB12±37%
Qwen3-Next-80B-A3B-InstructMoEQ4_K_S81.3B43.71 GiB3.00 GiB47.40 GiB0.60 GiB12±37%
CodeLlama-70b-Instruct-hfI1-Q4_K_S69.0B36.55 GiB10.00 GiB47.38 GiB0.62 GiB3±25%
CodeLlama-70b-Python-hfI1-Q4_K_S69.0B36.55 GiB10.00 GiB47.38 GiB0.62 GiB3±25%
Nous-Hermes-Llama2-70bI1-Q4_K_S69.0B36.55 GiB10.00 GiB47.38 GiB0.62 GiB3±25%
Midnight-Miqu-70B-v1.5I1-Q4_K_S69.0B36.55 GiB10.00 GiB47.38 GiB0.62 GiB3±25%
KafkaLM-70B-German-V0.1Q4_K_S69.0B36.55 GiB10.00 GiB47.38 GiB0.62 GiB3±25%
Mixtral_34Bx2_MoE_60BMoEQ5_K_M60.8B39.03 GiB7.50 GiB47.31 GiB0.69 GiB2±37%
GLM-4.5VMoEI1-Q2_K_S108B40.74 GiB5.75 GiB47.22 GiB0.78 GiB8±37%
llama2_70b_chat_uncensoredQ4_K_S69.0B36.39 GiB10.00 GiB47.21 GiB0.79 GiB3±25%
Xwin-LM-70b-V0.1Q4_K_S69.0B36.39 GiB10.00 GiB47.21 GiB0.79 GiB3±25%
Llama-2-70b-chat-hfQ4_K_S69.0B36.39 GiB10.00 GiB47.21 GiB0.79 GiB3±25%
MiniMax-M2.1-REAP-139B-A10BMoEI1-IQ2_S139B38.75 GiB7.75 GiB47.19 GiB0.81 GiB7±37%
m51Lab-MiniMax-M2.7-REAP-139B-A10BMoEI1-IQ2_S139B38.75 GiB7.75 GiB47.19 GiB0.81 GiB7±37%
Step-3.5-Flash-REAP-121B-A11BI1-IQ2_S121B33.40 GiB13.03 GiB47.16 GiB0.84 GiB3±25%
GPT-NeoX-20B-ErebusI1-Q5_K_S20.6B13.21 GiB33.00 GiB47.00 GiB1.00 GiB3±25%
DeepSeek-Coder-V2-Instruct-0724MoEIQ1_S236B44.14 GiB2.11 GiB46.99 GiB1.01 GiB13±37%
DeepSeek-V2.5MoEIQ1_S236B44.14 GiB2.11 GiB46.99 GiB1.01 GiB13±37%
DeepSeek-Coder-V2-InstructMoEIQ1_S236B44.14 GiB2.11 GiB46.99 GiB1.01 GiB13±37%
Assistant_Pepe_70BIQ4_XS70.6B36.11 GiB10.00 GiB46.93 GiB1.07 GiB3±25%
CalmeRys-78B-Orpo-v0.1I1-IQ3_M78.0B35.33 GiB10.75 GiB46.91 GiB1.09 GiB3±25%
●From the file◐Predictedwhat these mean

Speed is modeled, not measured: decode is memory-bandwidth bound, so tokens per second is bytes read per token against achievable bandwidth. Mixture-of-experts models carry a wider band because only the routed experts are read each step, and few have been measured publicly.

Questions people ask

What AI models can a AMD Ryzen AI Max+ 395 (Radeon 8060S) run?
2031 of 2118 indexed open-weight models fit a AMD Ryzen AI Max+ 395 (Radeon 8060S) at 32,768 context with f16 KV cache, the largest being command-a-plus-05-2026-bf16 at IQ1_S. That covers text, vision-language, image, video and speech models.
How much usable memory does a AMD Ryzen AI Max+ 395 (Radeon 8060S) actually have?
Its nameplate is 64 GB, but about 44.64 GiB is available to a model once driver and compositor overhead is accounted for, and only 48 GB of the pool can be allocated to the GPU at all.
Is a AMD Ryzen AI Max+ 395 (Radeon 8060S) fast for local AI?
Its memory bandwidth is 256 GB/s, and that figure — not teraflops — is what governs token generation speed. Capacity decides what you can run; bandwidth decides how fast it runs.