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. 2082 of 2118 indexed models fit at 128K 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 1788vision language 190audio tts 21image 2audio asr 39video 16embedding 26

What fits at 128K context

largest quantization that fits, per model · 2082 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
GLM-4.7-REAP-218B-A32BMoEIQ3_XXS218B82.27 GiB12.94 GiB95.94 GiB0.06 GiB3±37%
GLM-4.6-Derestricted-v3MoEIQ2_XXS357B82.26 GiB12.94 GiB95.94 GiB0.06 GiB4±37%
GLM-4.6MoEIQ2_XXS357B82.26 GiB12.94 GiB95.94 GiB0.06 GiB4±37%
HarmonicHarlequin_v5-20BI1-Q5_K_M33.3B21.98 GiB73.13 GiB95.85 GiB0.15 GiB1±25%
Qwen3-VL-235B-A22B-ThinkingMoEUD-IQ3_XXS236B88.22 GiB6.61 GiB95.57 GiB0.43 GiB5±37%
Qwen3-VL-235B-A22B-InstructMoEUD-IQ3_XXS236B88.15 GiB6.61 GiB95.49 GiB0.51 GiB5±37%
Devstral-2-123B-Instruct-2512Q5_K_M125B82.25 GiB12.38 GiB95.48 GiB0.52 GiB2±25%
Mistral-Medium-3.5-128BI1-Q5_K_M128B82.25 GiB12.38 GiB95.48 GiB0.52 GiB2±25%
XORTRON-NXTXPRTXXLI1-Q5_K_M128B82.25 GiB12.38 GiB95.48 GiB0.52 GiB2±25%
ERNIE-4.5-300B-A47B-PTUD-IQ1_M300B86.78 GiB7.59 GiB95.20 GiB0.80 GiB2±25%
MiMo-V2-FlashMoEKV unresolvedIQ2_M310B90.09 GiB4.22 GiB95.06 GiB0.94 GiB7±37%
Qwen3.5-122B-A10B-hereticMoEI1-Q6_K123B93.42 GiB0.84 GiB94.99 GiB1.01 GiB9±37%
MiMo-V2.5MoEKV unresolvedUD-IQ2_M311B89.93 GiB4.22 GiB94.89 GiB1.11 GiB7±37%
step-3.5-flashQ3_K_S199B80.06 GiB13.79 GiB94.58 GiB1.42 GiB2±25%
Step-3.5-Flash-REAP-121B-A11BI1-Q5_K_M121B79.79 GiB13.79 GiB94.31 GiB1.69 GiB2±25%
Dolphin3.0-R1-Mistral-24BF3223.6B87.82 GiB5.63 GiB94.26 GiB1.74 GiB2±25%
Dolphin3.0-Mistral-24BF3223.6B87.82 GiB5.63 GiB94.26 GiB1.74 GiB2±25%
Mistral-Small-24B-Instruct-2501-abliteratedF3223.6B87.82 GiB5.63 GiB94.26 GiB1.74 GiB2±25%
Mistral-Small-24B-Instruct-2501F3223.6B87.82 GiB5.63 GiB94.26 GiB1.74 GiB2±25%
Mistral-Small-4-119B-2603MoEUD-Q6_K119B92.60 GiB0.79 GiB94.12 GiB1.88 GiB9±37%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEQ5_K_L124B90.28 GiB3.09 GiB94.07 GiB1.93 GiB7±37%
Qwen3-235B-A22B-Thinking-2507MoEIQ3_XXS235B86.68 GiB6.61 GiB94.02 GiB1.98 GiB5±37%
Qwen3-235B-A22B-Instruct-2507MoEIQ3_XXS235B86.68 GiB6.61 GiB94.02 GiB1.98 GiB5±37%
Behemoth-X-123B-v2Q5_K_M123B80.55 GiB12.38 GiB93.78 GiB2.22 GiB2±25%
Mistral-Large-Instruct-2411Q5_K_M123B80.55 GiB12.38 GiB93.78 GiB2.22 GiB2±25%
Laguna-S-2.1MoEUD-Q6_K118B91.19 GiB1.73 GiB93.64 GiB2.36 GiB8±37%
DeepSeek-V4-FlashMoEQ2_K291B92.86 GiB0.02 GiB93.62 GiB2.38 GiB10±37%
Qwen3.5-397B-A17BMoEIQ1_M403B91.53 GiB1.05 GiB93.34 GiB2.66 GiB10±37%
MiniMax-M2.1MoEIQ3_XXS229B83.91 GiB8.72 GiB93.32 GiB2.68 GiB5±37%
MiniMax-M2MoEIQ3_XXS229B83.91 GiB8.72 GiB93.32 GiB2.68 GiB5±37%
Kimi-Linear-48B-A3B-InstructMoEBF1649.1B91.54 GiB1.07 GiB93.31 GiB2.69 GiB2±25%
command-a-plus-05-2026-bf16MoEQ3_K_S219B91.28 GiB1.24 GiB93.23 GiB2.77 GiB7±37%
Step-3.7-FlashIQ3_XXS201B78.38 GiB13.79 GiB92.90 GiB3.10 GiB2±25%
DeepSeek-Coder-V2-Instruct-0724MoEIQ3_XS236B89.69 GiB2.37 GiB92.80 GiB3.20 GiB7±37%
DeepSeek-V2.5MoEIQ3_XS236B89.69 GiB2.37 GiB92.80 GiB3.20 GiB7±37%
DeepSeek-Coder-V2-InstructMoEIQ3_XS236B89.69 GiB2.37 GiB92.80 GiB3.20 GiB7±37%
grok-2MoEUD-IQ1_S270B82.82 GiB9.00 GiB92.67 GiB3.33 GiB2±37%
GLM-4.7MoEUD-TQ1_0358B78.69 GiB12.94 GiB92.37 GiB3.63 GiB4±37%
dots.llm1.instMoEQ2_K_L143B56.67 GiB34.88 GiB92.27 GiB3.73 GiB2±37%
xLAM-8x7b-rMoEBF1646.7B86.99 GiB4.50 GiB92.23 GiB3.77 GiB3±37%
GLM-4.5MoEUD-TQ1_0358B78.54 GiB12.94 GiB92.22 GiB3.78 GiB4±37%
Nex-N2-ProMoEIQ1_M397B90.02 GiB1.05 GiB91.82 GiB4.18 GiB10±37%
Trinity-Large-PreviewMoEIQ2_XXS399B88.58 GiB2.33 GiB91.63 GiB4.37 GiB9±37%
Trinity-Large-TrueBaseMoEIQ2_XXS399B88.58 GiB2.33 GiB91.63 GiB4.37 GiB9±37%
Codestral-22B-v0.1F3222.2B82.88 GiB7.88 GiB91.51 GiB4.49 GiB2±25%
Qwen3-235B-A22B-abliteratedMoEI1-IQ3_XXS235B84.17 GiB6.61 GiB91.51 GiB4.49 GiB5±37%
Mixtral-8x22B-v0.1MoEF32141B82.84 GiB7.88 GiB91.48 GiB4.52 GiB3±37%
GLM-4.5-Air-REAP-82B-A12BMoEQ8_081.9B84.17 GiB6.47 GiB91.36 GiB4.64 GiB4±37%
MiniMax-M2.7-BF16-ultra-uncensored-hereticMoEI1-IQ3_XXS229B81.91 GiB8.72 GiB91.32 GiB4.68 GiB5±37%
MiniMax-M2.5MoEI1-IQ3_XXS229B81.91 GiB8.72 GiB91.32 GiB4.68 GiB5±37%
Gemma-4-Dark-Gemistry-31BQ6_K32.7B84.44 GiB5.95 GiB91.17 GiB4.83 GiB2±25%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQ6_K27.8B88.00 GiB2.25 GiB91.02 GiB4.98 GiB2±25%
MiniMax-M2.1-REAP-139B-A10BMoEI1-Q4_1139B81.21 GiB8.72 GiB90.61 GiB5.39 GiB5±37%
m51Lab-MiniMax-M2.7-REAP-139B-A10BMoEI1-Q4_1139B81.21 GiB8.72 GiB90.61 GiB5.39 GiB5±37%
Llama-4-Scout-17B-16E-InstructMoEKV unresolvedQ6_K_L109B83.13 GiB6.75 GiB90.61 GiB5.39 GiB5±37%
GLM-4.6-REAP-268B-A32BMoEUD-IQ1_M269B76.78 GiB12.94 GiB90.46 GiB5.54 GiB4±37%
gpt-oss-20b-hereticMoEQ5_120.9B88.57 GiB0.85 GiB90.10 GiB5.90 GiB5±37%
Huihui-gpt-oss-20b-BF16-abliteratedMoEQ5_120.9B88.57 GiB0.85 GiB90.10 GiB5.90 GiB5±37%
CalmeRys-78B-Orpo-v0.1Q8_078.0B77.16 GiB12.09 GiB90.08 GiB5.92 GiB2±25%
calme-2.3-rys-78bQ8_078.0B77.16 GiB12.09 GiB90.08 GiB5.92 GiB2±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?
2082 of 2118 indexed open-weight models fit a AMD Ryzen AI Max+ 395 (Radeon 8060S) at 131,072 context with q4_0 KV cache, the largest being GLM-4.7-REAP-218B-A32B at IQ3_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.