AMD · unified x86

AMD Ryzen AI Max+ 395 (Radeon 8060S)

AMD Ryzen AI Max+ 395 (Radeon 8060S) has 128 GB of VRAM at 256 GB/s — about 89.28 GiB usable after driver and compositor overhead. 2089 of 2118 indexed models fit at 16K context with q4_0 KV. Note only 96 GB of its 128 GB is allocatable to the GPU.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
128 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 1794vision language 191audio tts 21image 2audio asr 39video 16embedding 26

What fits at 16K context

largest quantization that fits, per model · 2089 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
MiMo-V2-FlashMoEKV unresolvedUD-IQ2_XXS310B94.58 GiB0.53 GiB95.86 GiB0.14 GiB9±37%
Laguna-S-2.1MoEQ6_K_L118B94.83 GiB0.25 GiB95.81 GiB0.19 GiB9±37%
DeepSeek-Coder-V2-Instruct-0724MoEIQ3_S236B94.70 GiB0.30 GiB95.73 GiB0.27 GiB9±37%
DeepSeek-V2.5MoEIQ3_S236B94.70 GiB0.30 GiB95.73 GiB0.27 GiB9±37%
DeepSeek-Coder-V2-InstructMoEIQ3_S236B94.70 GiB0.30 GiB95.73 GiB0.27 GiB9±37%
grok-2MoEQ2_K_L270B93.41 GiB1.13 GiB95.38 GiB0.62 GiB3±37%
Step-3.5-Flash-REAP-121B-A11BI1-Q6_K121B92.51 GiB1.98 GiB95.22 GiB0.78 GiB2±25%
MiniMax-M3MoEIQ1_M427B93.82 GiB0.53 GiB95.06 GiB0.94 GiB9±37%
dots.llm1.instMoEQ4_K_L143B89.95 GiB4.36 GiB95.04 GiB0.96 GiB6±37%
gemma-4-26B-A4B-it-Uncensored-MAXMoEF3225.8B94.02 GiB0.26 GiB94.96 GiB1.04 GiB2±25%
MiniMax-M2.1MoEI1-IQ3_M229B93.13 GiB1.09 GiB94.90 GiB1.10 GiB9±37%
MiniMax-M2.5MoEI1-IQ3_M229B93.13 GiB1.09 GiB94.90 GiB1.10 GiB9±37%
Mixtral-8x22B-Instruct-v0.1MoEQ5_K_M141B93.11 GiB0.98 GiB94.85 GiB1.15 GiB3±37%
Mixtral-8x22B-v0.1MoEQ5_K_M141B93.11 GiB0.98 GiB94.85 GiB1.15 GiB3±37%
Mixtral-8x22B-v0.1MoEQ5_K_M141B93.10 GiB0.98 GiB94.85 GiB1.15 GiB3±37%
GLM-4.6-REAP-268B-A32BMoEQ2_K_L269B92.11 GiB1.62 GiB94.47 GiB1.53 GiB6±37%
MiniMax-M2.7MoEQ3_K_S229B92.49 GiB1.09 GiB94.27 GiB1.73 GiB9±37%
Qwen3.5-122B-A10B-hereticMoEI1-Q6_K123B93.42 GiB0.11 GiB94.25 GiB1.75 GiB9±37%
Llama-4-Maverick-17B-128E-InstructMoEKV unresolvedTQ1_0402B92.59 GiB0.84 GiB94.16 GiB1.84 GiB11±37%
MiniMax-M2MoEQ3_K_S229B92.31 GiB1.09 GiB94.08 GiB1.92 GiB9±37%
GLM-4.5-Air-DerestrictedMoEQ6_K110B92.37 GiB0.81 GiB93.91 GiB2.09 GiB7±37%
GLM-4.5-AirMoEQ6_K110B92.37 GiB0.81 GiB93.91 GiB2.09 GiB7±37%
MiniMax-M2.1-REAP-139B-A10BMoEI1-Q5_K_M139B91.98 GiB1.09 GiB93.76 GiB2.24 GiB7±37%
m51Lab-MiniMax-M2.7-REAP-139B-A10BMoEI1-Q5_K_M139B91.98 GiB1.09 GiB93.76 GiB2.24 GiB7±37%
MiniMax-M2.7-BF16-ultra-uncensored-hereticMoEI1-IQ3_S229B91.94 GiB1.09 GiB93.71 GiB2.29 GiB9±37%
DeepSeek-V4-FlashMoEQ2_K291B92.86 GiB0.02 GiB93.62 GiB2.38 GiB10±37%
Mistral-Small-4-119B-2603MoEUD-Q6_K119B92.60 GiB0.10 GiB93.43 GiB2.57 GiB10±37%
Step-3.7-FlashUD-IQ4_NL201B90.63 GiB1.98 GiB93.34 GiB2.66 GiB2±25%
GLM-4.7MoEUD-IQ1_S358B90.50 GiB1.62 GiB92.86 GiB3.14 GiB7±37%
GLM-4.7-REAP-218B-A32BMoEQ3_K_S218B90.39 GiB1.62 GiB92.74 GiB3.26 GiB6±37%
GLM-4.5MoEUD-IQ1_S358B90.38 GiB1.62 GiB92.74 GiB3.26 GiB7±37%
GLM-4.6MoEUD-IQ1_S357B90.28 GiB1.62 GiB92.64 GiB3.36 GiB7±37%
Qwen3.5-397B-A17BMoEIQ1_M403B91.53 GiB0.13 GiB92.41 GiB3.59 GiB11±37%
Kimi-Linear-48B-A3B-InstructMoEBF1649.1B91.54 GiB0.13 GiB92.38 GiB3.62 GiB2±25%
command-a-plus-05-2026-bf16MoEQ3_K_S219B91.28 GiB0.26 GiB92.25 GiB3.75 GiB7±37%
Qwen3-235B-A22B-Instruct-2507MoEIQ3_XS235B90.25 GiB0.83 GiB91.80 GiB4.20 GiB7±37%
Qwen3-235B-A22B-Thinking-2507MoEIQ3_XS235B90.25 GiB0.83 GiB91.80 GiB4.20 GiB7±37%
Solar-Open2-250BMoEIQ3_XXS250B89.86 GiB0.84 GiB91.43 GiB4.57 GiB10±37%
Hermes-4-405BUD-IQ1_M406B88.23 GiB2.21 GiB91.42 GiB4.58 GiB2±25%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEQ5_K_L124B90.28 GiB0.39 GiB91.36 GiB4.64 GiB8±37%
MiMo-V2.5MoEKV unresolvedUD-IQ2_M311B89.93 GiB0.53 GiB91.20 GiB4.80 GiB9±37%
GLM-4.6VMoEQ6_K_L108B89.58 GiB0.81 GiB91.12 GiB4.88 GiB7±37%
Qwen3-235B-A22B-abliteratedMoEI1-IQ3_XS235B89.36 GiB0.83 GiB90.92 GiB5.08 GiB7±37%
Nex-N2-ProMoEIQ1_M397B90.02 GiB0.13 GiB90.90 GiB5.10 GiB11±37%
GLM-4.5VMoEI1-Q6_K108B89.28 GiB0.81 GiB90.81 GiB5.19 GiB7±37%
Hermes-3-Llama-3.1-405BIQ1_M406B87.08 GiB2.21 GiB90.27 GiB5.73 GiB2±25%
Trinity-Large-PreviewMoEIQ2_XXS399B88.58 GiB0.49 GiB89.79 GiB6.21 GiB11±37%
Trinity-Large-TrueBaseMoEIQ2_XXS399B88.58 GiB0.49 GiB89.79 GiB6.21 GiB11±37%
Qwen3-VL-235B-A22B-ThinkingMoEUD-IQ3_XXS236B88.22 GiB0.83 GiB89.78 GiB6.22 GiB7±37%
Qwen3-VL-235B-A22B-InstructMoEUD-IQ3_XXS236B88.15 GiB0.83 GiB89.71 GiB6.29 GiB7±37%
gpt-oss-20b-hereticMoEQ5_120.9B88.57 GiB0.11 GiB89.37 GiB6.63 GiB5±37%
Huihui-gpt-oss-20b-BF16-abliteratedMoEQ5_120.9B88.57 GiB0.11 GiB89.37 GiB6.63 GiB5±37%
Dolphin3.0-R1-Mistral-24BF3223.6B87.82 GiB0.70 GiB89.34 GiB6.66 GiB2±25%
Dolphin3.0-Mistral-24BF3223.6B87.82 GiB0.70 GiB89.34 GiB6.66 GiB2±25%
Mistral-Small-24B-Instruct-2501-abliteratedF3223.6B87.82 GiB0.70 GiB89.34 GiB6.66 GiB2±25%
Mistral-Small-24B-Instruct-2501F3223.6B87.82 GiB0.70 GiB89.34 GiB6.66 GiB2±25%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQ6_K27.8B88.00 GiB0.28 GiB89.05 GiB6.95 GiB2±25%
Hy3MoEIQ2_S299B86.43 GiB1.41 GiB88.57 GiB7.43 GiB8±37%
ERNIE-4.5-300B-A47B-PTUD-IQ1_M300B86.78 GiB0.95 GiB88.55 GiB7.45 GiB2±25%
gpt-oss-120b-Uncensored-xCloudMoEI1-Q5_K_M117B87.49 GiB0.17 GiB88.35 GiB7.65 GiB10±37%
●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?
2089 of 2118 indexed open-weight models fit a AMD Ryzen AI Max+ 395 (Radeon 8060S) at 16,384 context with q4_0 KV cache, the largest being MiMo-V2-Flash at UD-IQ2_XXS. 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 128 GB, but about 89.28 GiB is available to a model once driver and compositor overhead is accounted for, and only 96 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.