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 64K 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
text 1788vision language 190audio tts 21image 2audio asr 39video 16embedding 26

What fits at 64K context

largest quantization that fits, per model · 2082 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
GLM-4.6-REAP-268B-A32BMoEUD-IQ2_XXS269B83.03 GiB12.22 GiB95.99 GiB0.01 GiB4±37%
Solar-Open2-250BMoEQ2_K250B88.87 GiB6.38 GiB95.97 GiB0.03 GiB6±37%
GLM-4.5MoEIQ2_XXS358B83.01 GiB12.22 GiB95.97 GiB0.03 GiB4±37%
MiniMax-M2.7-BF16-ultra-uncensored-hereticMoEI1-IQ3_XS229B86.92 GiB8.23 GiB95.84 GiB0.16 GiB5±37%
MiniMax-M2.1MoEI1-IQ3_XS229B86.92 GiB8.23 GiB95.84 GiB0.16 GiB5±37%
MiniMax-M2.5MoEI1-IQ3_XS229B86.92 GiB8.23 GiB95.84 GiB0.16 GiB5±37%
GLM-4.7MoEIQ2_XXS358B82.69 GiB12.22 GiB95.65 GiB0.35 GiB4±37%
Step-3.7-FlashQ3_K_S201B81.55 GiB13.30 GiB95.58 GiB0.42 GiB2±25%
dots.llm1.instMoEIQ3_XS143B61.71 GiB32.94 GiB95.37 GiB0.63 GiB2±37%
HarmonicHarlequin_v5-20BI1-Q6_K33.3B25.46 GiB69.06 GiB95.26 GiB0.74 GiB2±25%
GLM-4.7-REAP-218B-A32BMoEIQ3_XXS218B82.27 GiB12.22 GiB95.23 GiB0.77 GiB3±37%
GLM-4.6-Derestricted-v3MoEIQ2_XXS357B82.26 GiB12.22 GiB95.22 GiB0.78 GiB4±37%
GLM-4.6MoEIQ2_XXS357B82.26 GiB12.22 GiB95.22 GiB0.78 GiB4±37%
Qwen3-VL-235B-A22B-ThinkingMoEUD-IQ3_XXS236B88.22 GiB6.24 GiB95.20 GiB0.80 GiB5±37%
Qwen3-VL-235B-A22B-InstructMoEUD-IQ3_XXS236B88.15 GiB6.24 GiB95.12 GiB0.88 GiB5±37%
Qwen3.5-122B-A10B-hereticMoEI1-Q6_K123B93.42 GiB0.80 GiB94.94 GiB1.06 GiB9±37%
MiMo-V2-FlashMoEKV unresolvedIQ2_M310B90.09 GiB3.98 GiB94.83 GiB1.17 GiB7±37%
Devstral-2-123B-Instruct-2512Q5_K_M125B82.25 GiB11.69 GiB94.80 GiB1.20 GiB2±25%
Mistral-Medium-3.5-128BI1-Q5_K_M128B82.25 GiB11.69 GiB94.80 GiB1.20 GiB2±25%
XORTRON-NXTXPRTXXLI1-Q5_K_M128B82.25 GiB11.69 GiB94.80 GiB1.20 GiB2±25%
ERNIE-4.5-300B-A47B-PTUD-IQ1_M300B86.78 GiB7.17 GiB94.78 GiB1.22 GiB2±25%
MiMo-V2.5MoEKV unresolvedUD-IQ2_M311B89.93 GiB3.98 GiB94.66 GiB1.34 GiB7±37%
step-3.5-flashQ3_K_S199B80.06 GiB13.30 GiB94.09 GiB1.91 GiB2±25%
Mistral-Small-4-119B-2603MoEUD-Q6_K119B92.60 GiB0.75 GiB94.08 GiB1.92 GiB9±37%
Dolphin3.0-R1-Mistral-24BF3223.6B87.82 GiB5.31 GiB93.95 GiB2.05 GiB2±25%
Dolphin3.0-Mistral-24BF3223.6B87.82 GiB5.31 GiB93.95 GiB2.05 GiB2±25%
Mistral-Small-24B-Instruct-2501-abliteratedF3223.6B87.82 GiB5.31 GiB93.95 GiB2.05 GiB2±25%
Mistral-Small-24B-Instruct-2501F3223.6B87.82 GiB5.31 GiB93.95 GiB2.05 GiB2±25%
NVIDIA-Nemotron-3-Super-120B-A12B-BF16MoEQ5_K_L124B90.28 GiB2.92 GiB93.89 GiB2.11 GiB7±37%
Step-3.5-Flash-REAP-121B-A11BI1-Q5_K_M121B79.79 GiB13.30 GiB93.81 GiB2.19 GiB2±25%
Qwen3-235B-A22B-Thinking-2507MoEIQ3_XXS235B86.68 GiB6.24 GiB93.66 GiB2.34 GiB5±37%
Qwen3-235B-A22B-Instruct-2507MoEIQ3_XXS235B86.68 GiB6.24 GiB93.66 GiB2.34 GiB5±37%
DeepSeek-V4-FlashMoEQ2_K291B92.86 GiB0.03 GiB93.64 GiB2.36 GiB10±37%
Laguna-S-2.1MoEUD-Q6_K118B91.19 GiB1.67 GiB93.58 GiB2.42 GiB8±37%
Qwen3.5-397B-A17BMoEIQ1_M403B91.53 GiB1.00 GiB93.28 GiB2.72 GiB10±37%
command-a-plus-05-2026-bf16MoEQ3_K_S219B91.28 GiB1.29 GiB93.27 GiB2.73 GiB7±37%
Kimi-Linear-48B-A3B-InstructMoEBF1649.1B91.54 GiB1.01 GiB93.25 GiB2.75 GiB2±25%
Behemoth-X-123B-v2Q5_K_M123B80.55 GiB11.69 GiB93.09 GiB2.91 GiB2±25%
Mistral-Large-Instruct-2411Q5_K_M123B80.55 GiB11.69 GiB93.09 GiB2.91 GiB2±25%
MiniMax-M2MoEIQ3_XXS229B83.91 GiB8.23 GiB92.83 GiB3.17 GiB5±37%
DeepSeek-Coder-V2-Instruct-0724MoEIQ3_XS236B89.69 GiB2.24 GiB92.67 GiB3.33 GiB7±37%
DeepSeek-V2.5MoEIQ3_XS236B89.69 GiB2.24 GiB92.67 GiB3.33 GiB7±37%
DeepSeek-Coder-V2-InstructMoEIQ3_XS236B89.69 GiB2.24 GiB92.67 GiB3.33 GiB7±37%
grok-2MoEUD-IQ1_S270B82.82 GiB8.50 GiB92.17 GiB3.83 GiB2±37%
xLAM-8x7b-rMoEBF1646.7B86.99 GiB4.25 GiB91.98 GiB4.02 GiB3±37%
Nex-N2-ProMoEIQ1_M397B90.02 GiB1.00 GiB91.77 GiB4.23 GiB10±37%
Trinity-Large-PreviewMoEIQ2_XXS399B88.58 GiB2.41 GiB91.71 GiB4.29 GiB9±37%
Trinity-Large-TrueBaseMoEIQ2_XXS399B88.58 GiB2.41 GiB91.71 GiB4.29 GiB9±37%
Gemma-4-Dark-Gemistry-31BQ6_K32.7B84.44 GiB5.94 GiB91.16 GiB4.84 GiB2±25%
Qwen3-235B-A22B-abliteratedMoEI1-IQ3_XXS235B84.17 GiB6.24 GiB91.14 GiB4.86 GiB5±37%
Codestral-22B-v0.1F3222.2B82.88 GiB7.44 GiB91.08 GiB4.92 GiB2±25%
Mixtral-8x22B-v0.1MoEF32141B82.84 GiB7.44 GiB91.04 GiB4.96 GiB3±37%
GLM-4.5-Air-REAP-82B-A12BMoEQ8_081.9B84.17 GiB6.11 GiB91.00 GiB5.00 GiB5±37%
Qwen3.6-27B-Fable-Fusion-711-Uncensored-Heretic-NM-DAU-MTPQ6_K27.8B88.00 GiB2.13 GiB90.89 GiB5.11 GiB2±25%
Llama-4-Scout-17B-16E-InstructMoEKV unresolvedQ6_K_L109B83.13 GiB6.38 GiB90.23 GiB5.77 GiB5±37%
MiniMax-M2.1-REAP-139B-A10BMoEI1-Q4_1139B81.21 GiB8.23 GiB90.13 GiB5.87 GiB5±37%
m51Lab-MiniMax-M2.7-REAP-139B-A10BMoEI1-Q4_1139B81.21 GiB8.23 GiB90.13 GiB5.87 GiB5±37%
gpt-oss-20b-hereticMoEQ5_120.9B88.57 GiB0.81 GiB90.06 GiB5.94 GiB5±37%
Huihui-gpt-oss-20b-BF16-abliteratedMoEQ5_120.9B88.57 GiB0.81 GiB90.06 GiB5.94 GiB5±37%
CalmeRys-78B-Orpo-v0.1Q8_078.0B77.16 GiB11.42 GiB89.41 GiB6.59 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 65,536 context with q8_0 KV cache, the largest being GLM-4.6-REAP-268B-A32B 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.