AMD · unified x86

AMD Ryzen AI Max 385 (Radeon 8050S)

AMD Ryzen AI Max 385 (Radeon 8050S) has 32 GB of VRAM at 256 GB/s — about 22.32 GiB usable after driver and compositor overhead. 1984 of 2118 indexed models fit at 8K 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 1704vision language 176audio tts 21image 2video 16audio asr 39embedding 26

What fits at 8K context

largest quantization that fits, per model · 1984 of 2118 indexed
ModelBest quantParamsWeights●KV●Total in memory◐Headroom◐tok/s◐
solar-pro-preview-instructKV unresolvedQ8_022.1B21.91 GiB1.33 GiB24.00 GiB0.00 GiB6±25%
Qwen3-Coder-Next-REAMMoEI1-IQ3_XS60.3B23.19 GiB0.10 GiB23.98 GiB0.02 GiB37±37%
CallerQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Dumpling-Qwen2.5-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OREAL-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
openhands-lm-32b-v0.1Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
LongWriter-Zero-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OpenCodeReasoning-Nemotron-32B-IOIQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-Coder-32B-Instruct-abliteratedQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OlympicCoder-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OpenCodeReasoning-Nemotron-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OpenThinker-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
QwQ-32B-ArliAI-RpR-v4Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-Coder-32B-InstructQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
QwQ-32B-abliteratedQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
OpenThinker2-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
INTELLECT-2Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-32B-InstructQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
QwQ-32B-PreviewQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-Coder-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-32b-RP-InkQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
TinyR1-32B-PreviewQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
deepseek-r1-qwen-2.5-32B-ablatedQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Rombos-LLM-V2.5-Qwen-32bQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
DeepSeek-R1-Distill-Qwen-32B-abliteratedQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-32B-ArliAI-RPMax-v1.3Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
DeepSeek-R1-Distill-Qwen-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen2.5-VL-32B-InstructQ5_K_L33.5B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
EVA-Qwen2.5-32B-v0.2Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
EVA-Qwen2.5-32B-v0.1Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
cogito-v1-preview-qwen-32BQ5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
QwQ-32B-Snowdrop-v0Q5_K_L32.8B22.11 GiB1.06 GiB23.97 GiB0.03 GiB6±25%
Qwen3.5-122B-A10B-hereticMoEI1-IQ1_S123B23.13 GiB0.10 GiB23.96 GiB0.04 GiB35±37%
command-r-35b-writer-v2I1-Q3_K_L35.0B17.83 GiB5.31 GiB23.95 GiB0.05 GiB6±25%
spoomplesmaxx-v2.1-30BI1-Q6_K28.9B22.06 GiB1.06 GiB23.93 GiB0.07 GiB6±25%
Huihui-granite-4.1-30b-abliteratedI1-Q6_K28.9B22.06 GiB1.06 GiB23.93 GiB0.07 GiB6±25%
granite-4.1-30b-hereticI1-Q6_K28.9B22.06 GiB1.06 GiB23.93 GiB0.07 GiB6±25%
granite-4.1-30bQ6_K28.9B22.06 GiB1.06 GiB23.93 GiB0.07 GiB6±25%
GLM-4.7-FlashMoEQ6_K31.2B23.00 GiB0.22 GiB23.93 GiB0.07 GiB26±37%
Nemotron-Cascade-2-30B-A3BMoEQ4_K_M31.6B23.03 GiB0.22 GiB23.92 GiB0.08 GiB29±37%
OpenBuddy-R1-0528-Distill-Qwen3-32B-Preview0-QATQ5_K_L32.8B22.07 GiB1.06 GiB23.92 GiB0.08 GiB6±25%
KAT-DevQ5_K_L32.8B22.07 GiB1.06 GiB23.92 GiB0.08 GiB6±25%
Qwen3-VL-32B-InstructQ5_K_L33.4B22.07 GiB1.06 GiB23.92 GiB0.08 GiB6±25%
DeepSWE-PreviewQ5_K_L32.8B22.07 GiB1.06 GiB23.92 GiB0.08 GiB6±25%
Bielik-11B-v2.3-InstructQ8_011.2B22.34 GiB0.83 GiB23.90 GiB0.10 GiB6±25%
deepseek-llm-67b-chatI1-IQ2_M67.4B21.53 GiB1.58 GiB23.90 GiB0.10 GiB6±25%
deepseek-llm-67b-baseI1-IQ2_M67.4B21.53 GiB1.58 GiB23.90 GiB0.10 GiB6±25%
Nemotron-Labs-Audex-30B-A3BQ4_K_M32.0B23.16 GiB0.00 GiB23.90 GiB0.10 GiB6±25%
openbuddy-deepseek-67b-v15.3-4kI1-IQ2_M67.4B21.53 GiB1.58 GiB23.90 GiB0.10 GiB6±25%
Yi-34B-200K-DARE-megamerge-v8I1-Q5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
dolphin-2.9.1-yi-1.5-34b-hereticQ5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
dolphin-2.9.1-yi-1.5-34bI1-Q5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
OrionStar-Yi-34B-Chat-LlamaI1-Q5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Yi-34B-200K-LlamafiedI1-Q5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Yi-1.5-34BQ5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Nous-Hermes-2-Yi-34BQ5_034.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Merged-RP-Stew-V2-34BI1-Q5_K_S34.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Capybara-Tess-Yi-34B-200KQ5_034.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
Nous-Capybara-limarpv3-34BQ5_034.4B22.08 GiB1.00 GiB23.86 GiB0.14 GiB6±25%
GLM-4.7-Flash-DerestrictedMoEI1-Q6_K31.2B22.92 GiB0.22 GiB23.85 GiB0.15 GiB26±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 385 (Radeon 8050S) run?
1984 of 2118 indexed open-weight models fit a AMD Ryzen AI Max 385 (Radeon 8050S) at 8,192 context with q8_0 KV cache, the largest being solar-pro-preview-instruct at Q8_0. That covers text, vision-language, image, video and speech models.
How much usable memory does a AMD Ryzen AI Max 385 (Radeon 8050S) 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 385 (Radeon 8050S) 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.