AMD · unified x86

AMD Ryzen AI Max+ 395 (Radeon 8060S)

AMD Ryzen AI Max+ 395 (Radeon 8060S) has 32 GB of VRAM at 256 GB/s — about 22.32 GiB usable after driver and compositor overhead. 1996 of 2118 indexed models fit at 4K context with q8_0 KV. Note only 24 GB of its 32 GB is allocatable to the GPU.

Spec sheet· bandwidth, theoreticalFrom the file· fit from summed bytesPredicted· speed
Memory
32 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 1714vision language 178audio tts 21image 2video 16audio asr 39embedding 26

What fits at 4K context

largest quantization that fits, per model · 1996 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
Le-Chaton-Slim-23BMoEQ8_023.3B23.07 GiB0.22 GiB24.00 GiB0.00 GiB13±37%
Qwen-AgentWorld-35B-A3BMoEUD-Q5_K_S34.7B23.23 GiB0.04 GiB23.97 GiB0.03 GiB36±37%
Ornith-1.0-35BMoEUD-Q5_K_S34.7B23.23 GiB0.04 GiB23.97 GiB0.03 GiB36±37%
Qwen3.5-40B-Claude-4.6-Opus-Deckard-Heretic-Uncensored-ThinkingI1-Q4_139.5B23.00 GiB0.20 GiB23.96 GiB0.04 GiB6±25%
Qwen3.5-40B-RoughHouse-Claude-4.6-Opus-Polar-Deckard-Uncensored-Heretic-ThinkingI1-Q4_139.5B23.00 GiB0.20 GiB23.96 GiB0.04 GiB6±25%
Maenad-70BI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
DeepSeek-R1-Distill-Llama-70B-Uncensored-v2-Unbiased-ReasonerI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Rombos-LLM-70b-Llama-3.3I1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
L3.3-Electra-R1-70bI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
L3.3-70B-Magnum-v4-SEIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Latxa-Llama-3.1-70B-Instruct-v2I1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.3_70_b_uncensored_continuedI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.3-70B-Instruct-abliteratedI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
grok-oss-Revenant-70BI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.1-Nemotron-70B-Instruct-HFI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
L3.3-70B-Euryale-v2.3I1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Hermes-3-Llama-3.1-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Hermes-4-70B-hereticI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.3-70B-InstructIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.1-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Anubis-70B-v1.2IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Hermes-4-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Golem-70B-v1bI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
DeepSeek-R1-Distill-Llama-70B-abliteratedI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
DeepSeek-R1-Distill-Llama-70B-hereticI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
DeepSeek-R1-Distill-Llama-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Legion-V2.1-LLaMa-70BI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Assistant_Pepe_70BI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Tess-R1-Limerick-Llama-3.1-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
SEMIKONG-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
functionary-medium-v3.2KV unresolvedIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Llama-3.1-WhiteRabbitNeo-2-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Infinity-Instruct-7M-Gen-Llama3_1-70BI1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
New-Dawn-Llama-3-70B-32K-v1.0I1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Meta-Llama-3-70B-Instruct-abliterated-v3.5I1-IQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Athene-70BIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
L3.3-70B-Magnum-DiamondIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Meta-Llama-3-70B-InstructIQ2_M70.6B22.46 GiB0.66 GiB23.95 GiB0.05 GiB6±25%
Nemotron-Cascade-2-30B-A3BMoEQ4_K_L31.6B23.15 GiB0.11 GiB23.94 GiB0.06 GiB30±37%
Yi-34B-200K-DARE-megamerge-v8I1-Q5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
dolphin-2.9.1-yi-1.5-34b-hereticQ5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
dolphin-2.9.1-yi-1.5-34bI1-Q5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
OrionStar-Yi-34B-Chat-LlamaI1-Q5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Yi-34B-200K-LlamafiedI1-Q5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Yi-1.5-34BQ5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Nous-Hermes-2-Yi-34BQ5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Merged-RP-Stew-V2-34BI1-Q5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Capybara-Tess-Yi-34B-200KQ5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Nous-Capybara-limarpv3-34BQ5_K_M34.4B22.65 GiB0.50 GiB23.93 GiB0.07 GiB6±25%
Qwen3-Coder-Next-REAMMoEI1-IQ3_XS60.3B23.19 GiB0.05 GiB23.93 GiB0.07 GiB38±37%
Qwen3.5-122B-A10B-hereticMoEI1-IQ1_S123B23.13 GiB0.05 GiB23.91 GiB0.09 GiB36±37%
Hy-MT2-30B-A3BMoEQ6_K30.1B23.01 GiB0.20 GiB23.91 GiB0.09 GiB26±37%
Nemotron-Labs-Audex-30B-A3BQ4_K_M32.0B23.16 GiB0.00 GiB23.90 GiB0.10 GiB6±25%
KrakenSakura-Maelstrom-12B-v1F1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
Mistral-Nemo-Base-2407F1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
Vikhr-Nemo-12B-Instruct-R-21-09-24F1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
Wayfarer-2-12BBF1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
Wayfarer-12BF1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
writing-roleplay-20k-context-nemo-12b-v1.0F1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±25%
Muse-12BBF1612.2B22.82 GiB0.33 GiB23.90 GiB0.10 GiB6±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?
1996 of 2118 indexed open-weight models fit a AMD Ryzen AI Max+ 395 (Radeon 8060S) at 4,096 context with q8_0 KV cache, the largest being Le-Chaton-Slim-23B at Q8_0. 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 32 GB, but about 22.32 GiB is available to a model once driver and compositor overhead is accounted for, and only 24 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.