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. 2087 of 2118 indexed models fit at 32K context with q8_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
vision language 190text 1793audio tts 21image 2audio asr 39video 16embedding 26

What fits at 32K context

largest quantization that fits, per model · 2087 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
gemma-4-26B-A4B-it-Uncensored-MAXMoEF3225.8B94.02 GiB0.82 GiB95.52 GiB0.48 GiB2±25%
Step-3.7-FlashQ3_K_M201B87.36 GiB6.92 GiB95.01 GiB0.99 GiB2±25%
Mixtral-8x22B-Instruct-v0.1MoEQ5_K_S141B90.32 GiB3.72 GiB94.80 GiB1.20 GiB3±37%
Mixtral-8x22B-v0.1MoEQ5_K_S141B90.31 GiB3.72 GiB94.79 GiB1.21 GiB3±37%
Mixtral-8x22B-v0.1MoEQ5_K_S141B90.31 GiB3.72 GiB94.79 GiB1.21 GiB3±37%
dots.llm1.instMoEQ4_0143B77.44 GiB16.47 GiB94.63 GiB1.37 GiB3±37%
Qwen3.5-122B-A10B-hereticMoEI1-Q6_K123B93.42 GiB0.40 GiB94.55 GiB1.45 GiB9±37%
Llama-3_1-Nemotron-51B-InstructQ8_051.5B50.97 GiB42.50 GiB94.31 GiB1.69 GiB2±25%
GLM-4.6-REAP-268B-A32BMoEUD-IQ2_M269B87.26 GiB6.11 GiB94.11 GiB1.89 GiB5±37%
Qwen3-235B-A22B-Instruct-2507MoEIQ3_XS235B90.25 GiB3.12 GiB94.10 GiB1.90 GiB6±37%
Qwen3-235B-A22B-Thinking-2507MoEIQ3_XS235B90.25 GiB3.12 GiB94.10 GiB1.90 GiB6±37%
MiniMax-M2.1-REAP-139B-A10BMoEI1-Q5_K_S139B89.27 GiB4.12 GiB94.07 GiB1.93 GiB6±37%
m51Lab-MiniMax-M2.7-REAP-139B-A10BMoEI1-Q5_K_S139B89.27 GiB4.12 GiB94.07 GiB1.93 GiB6±37%
Solar-Open2-250BMoEIQ3_XXS250B89.86 GiB3.19 GiB93.77 GiB2.23 GiB8±37%
Mistral-Small-4-119B-2603MoEUD-Q6_K119B92.60 GiB0.37 GiB93.71 GiB2.29 GiB9±37%
DeepSeek-V4-FlashMoEQ2_K291B92.86 GiB0.03 GiB93.64 GiB2.36 GiB10±37%
MiniMax-M2.7MoEIQ3_XXS229B88.65 GiB4.12 GiB93.45 GiB2.55 GiB7±37%
GLM-4.6VMoEQ6_K_L108B89.58 GiB3.05 GiB93.37 GiB2.63 GiB6±37%
Qwen3-235B-A22B-abliteratedMoEI1-IQ3_XS235B89.36 GiB3.12 GiB93.22 GiB2.78 GiB6±37%
grok-2MoEUD-IQ1_M270B88.02 GiB4.25 GiB93.11 GiB2.89 GiB3±37%
GLM-4.5VMoEI1-Q6_K108B89.28 GiB3.05 GiB93.06 GiB2.94 GiB6±37%
MiMo-V2-FlashMoEKV unresolvedIQ2_M310B90.09 GiB1.99 GiB92.83 GiB3.17 GiB8±37%
Laguna-S-2.1MoEUD-Q6_K118B91.19 GiB0.87 GiB92.78 GiB3.22 GiB8±37%
Qwen3.5-397B-A17BMoEIQ1_M403B91.53 GiB0.50 GiB92.78 GiB3.22 GiB10±37%
Kimi-Linear-48B-A3B-InstructMoEBF1649.1B91.54 GiB0.50 GiB92.75 GiB3.25 GiB2±25%
command-a-plus-05-2026-bf16MoEQ3_K_S219B91.28 GiB0.76 GiB92.74 GiB3.26 GiB7±37%
Llama-3_3-Nemotron-Super-49B-v1_5Q8_049.9B49.36 GiB42.50 GiB92.70 GiB3.30 GiB2±25%
Valkyrie-49B-v2.1Q8_049.9B49.36 GiB42.50 GiB92.70 GiB3.30 GiB2±25%
Llama-3_3-Nemotron-Super-49B-v1Q8_049.9B49.36 GiB42.50 GiB92.70 GiB3.30 GiB2±25%
MiMo-V2.5MoEKV unresolvedUD-IQ2_M311B89.93 GiB1.99 GiB92.67 GiB3.33 GiB8±37%
Mistral-Medium-3.5-128BQ5_K_L128B85.78 GiB5.84 GiB92.48 GiB3.52 GiB2±25%
Hy3MoEIQ2_S299B86.43 GiB5.31 GiB92.48 GiB3.52 GiB6±37%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEQ5_K_L124B90.28 GiB1.46 GiB92.43 GiB3.57 GiB8±37%
GLM-4.7-REAP-218B-A32BMoEIQ3_XS218B85.49 GiB6.11 GiB92.34 GiB3.66 GiB5±37%
MiniMax-M2.1MoEIQ3_XS229B87.32 GiB4.12 GiB92.13 GiB3.87 GiB7±37%
MiniMax-M2MoEIQ3_XS229B87.32 GiB4.12 GiB92.13 GiB3.87 GiB7±37%
Qwen3-VL-235B-A22B-ThinkingMoEUD-IQ3_XXS236B88.22 GiB3.12 GiB92.08 GiB3.92 GiB6±37%
Qwen3-VL-235B-A22B-InstructMoEUD-IQ3_XXS236B88.15 GiB3.12 GiB92.00 GiB4.00 GiB6±37%
MiniMax-M2.7-BF16-ultra-uncensored-hereticMoEI1-IQ3_XS229B86.92 GiB4.12 GiB91.72 GiB4.28 GiB7±37%
MiniMax-M2.5MoEI1-IQ3_XS229B86.92 GiB4.12 GiB91.72 GiB4.28 GiB7±37%
step-3.5-flashQ3_K_M199B84.04 GiB6.92 GiB91.69 GiB4.31 GiB2±25%
DeepSeek-Coder-V2-Instruct-0724MoEIQ3_XS236B89.69 GiB1.12 GiB91.55 GiB4.45 GiB8±37%
DeepSeek-V2.5MoEIQ3_XS236B89.69 GiB1.12 GiB91.55 GiB4.45 GiB8±37%
DeepSeek-Coder-V2-InstructMoEIQ3_XS236B89.69 GiB1.12 GiB91.55 GiB4.45 GiB8±37%
Dolphin3.0-R1-Mistral-24BF3223.6B87.82 GiB2.66 GiB91.30 GiB4.70 GiB2±25%
Dolphin3.0-Mistral-24BF3223.6B87.82 GiB2.66 GiB91.30 GiB4.70 GiB2±25%
Mistral-Small-24B-Instruct-2501-abliteratedF3223.6B87.82 GiB2.66 GiB91.30 GiB4.70 GiB2±25%
Mistral-Small-24B-Instruct-2501F3223.6B87.82 GiB2.66 GiB91.30 GiB4.70 GiB2±25%
Nex-N2-ProMoEIQ1_M397B90.02 GiB0.50 GiB91.27 GiB4.73 GiB11±37%
ERNIE-4.5-300B-A47B-PTUD-IQ1_M300B86.78 GiB3.59 GiB91.19 GiB4.81 GiB2±25%
Trinity-Large-PreviewMoEIQ2_XXS399B88.58 GiB1.42 GiB90.72 GiB5.28 GiB10±37%
Trinity-Large-TrueBaseMoEIQ2_XXS399B88.58 GiB1.42 GiB90.72 GiB5.28 GiB10±37%
Hermes-4-405BUD-IQ1_S406B81.23 GiB8.37 GiB90.58 GiB5.42 GiB2±25%
GLM-4.5MoEIQ2_XXS358B83.01 GiB6.11 GiB89.86 GiB6.14 GiB5±37%
xLAM-8x7b-rMoEBF1646.7B86.99 GiB2.13 GiB89.86 GiB6.14 GiB3±37%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQ6_K27.8B88.00 GiB1.06 GiB89.83 GiB6.17 GiB2±25%
gpt-oss-20b-hereticMoEQ5_120.9B88.57 GiB0.41 GiB89.66 GiB6.34 GiB5±37%
Huihui-gpt-oss-20b-BF16-abliteratedMoEQ5_120.9B88.57 GiB0.41 GiB89.66 GiB6.34 GiB5±37%
GLM-4.7MoEIQ2_XXS358B82.69 GiB6.11 GiB89.54 GiB6.46 GiB5±37%
GLM-4.6-Derestricted-v3MoEIQ2_XXS357B82.26 GiB6.11 GiB89.11 GiB6.89 GiB5±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?
2087 of 2118 indexed open-weight models fit a AMD Ryzen AI Max+ 395 (Radeon 8060S) at 32,768 context with q8_0 KV cache, the largest being gemma-4-26B-A4B-it-Uncensored-MAX at F32. 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.