# What Models Can Ryzen AI Max+ 395 Run? Memory Fit Calculator > Estimate model fit on Ryzen AI Max+ 395 at 64GB, 96GB or 128GB across parameter count, quantization, context and sessions. - Best used for: Use when the user needs to calculate, size, estimate, or plan: What Models Can Ryzen AI Max+ 395 Run? Memory Fit Calculator - Canonical: https://cloudzat.com/what-models-can-ryzen-ai-max-395-run/ - Published: 2026-09-09 - Updated: 2026-09-09 - Author: Kayla Idayi - Site: https://cloudzat.com/ - LLM index: https://cloudzat.com/llms.txt ## Content [Home](https://cloudzat.com/)/ [Local AI](https://cloudzat.com/local-ai/)/ What Models Can Ryzen AI Max+ 395 Run? Memory Fit Calculator Ryzen AI Max+ 395 model-fit calculator # What Models Can Ryzen AI Max+ 395 Run? Memory Fit Calculator The Ryzen AI Max+ 395 can support unusually large local models because its Radeon 8060S shares a memory pool that can reach 128GB. The practical limit depends on quantization, context length, runtime overhead and how much memory must remain for the rest of the system. This calculator estimates memory fit, not tokens per second. A model that fits can still run slowly or require a specific software path. Quick answer A 128GB Ryzen AI Max+ 395 can accommodate many quantized models that exceed consumer GPU VRAM, including large 70B-class workloads and some substantially larger models. Use the calculator with your quantization and context target rather than relying on parameter count alone. **Before you buy** Leave headroom. Planning to consume every advertised gigabyte makes the system fragile once the operating system, context cache and other applications are active. CPU**16 cores / 32 threads**Zen 5, up to 5.1GHz boost Memory**Up to 128GB**256-bit LPDDR5X-8000 shared pool Graphics**Radeon 8060S**40 graphics CUs, up to 2.9GHz Power range**45-120W**Finished-system tuning matters Interactive fit tool ## Ryzen AI Max+ 395 model-fit calculator Estimate weight footprint, context reserve and working headroom using the same Cloudzat model-fit engine as the Ryzen AI Max+ PRO 495 calculator. Memory configuration64GB96GB128GB Model size7B model14B model27B model32B model70B model109B model120B model235B modelCustom Custom parameters (billions) QuantizationFP168-bit6-bit5-bit4-bit Context target8K32K64K128K Concurrent sessions124 **Choose the options that match your workload.** The result will focus on the trade-off that matters most for this decision. Amazon Live ## Systems for larger local models The 128GB systems below offer the widest model-fit envelope in the 395 family. Loading current Amazon buying options… Amazon prices, sellers and availability can change. Confirm the exact configuration and checkout total before buying. Buying decision ## Buying decision | Question | Best fit | Reason | | --- | --- | --- | | Comfortable estimate | Proceed to runtime testing | The memory plan retains useful reserve after estimated weights and context. | | Constrained estimate | Reduce context or choose more memory | The model may fit but leaves little room for the rest of the workload. | | Tight estimate | Use a smaller quantization or higher tier | Running at the edge of memory is unlikely to be robust. | | Does not fit | Choose a smaller model or larger-memory platform | Capacity must be solved before speed matters. | Before you buy ## Four checks that matter ### Memory capacity Confirm the exact soldered memory configuration because the wide LPDDR5X pool is not a normal user-upgradeable DIMM path. ### Power and cooling Check sustained power mode and cooling, not only the processor badge; long AI and creator workloads can separate otherwise similar systems. ### Ports and expansion Match storage, networking, USB4 and PCIe needs to the finished machine before choosing the lowest listed price. ### Software fit Verify the runtime and acceleration path used by your actual models or applications before assuming shared-memory capacity guarantees performance. ## How the model-fit estimate should be read The calculator estimates whether model weights plus a context allowance and runtime reserve fit within a chosen 395 memory profile. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. It is a planning tool rather than a benchmark; exact runtimes can allocate memory differently. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Parameters are not the same as memory use A 70B model does not require exactly 70GB because bit depth, metadata and runtime structures alter the weight footprint. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Quantization is therefore one of the largest inputs in any practical fit estimate. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## FP16 is a very different target from four-bit Full-precision weights can exceed the useful memory budget quickly, while four-bit quantization can reduce the model footprint dramatically. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Choose the format you will actually run rather than comparing parameter count alone. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Context is a second major memory consumer KV cache and related runtime state grow with context and can turn a comfortable short-context model into a tight long-context workload. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. If long documents or coding repositories matter, test the context target you expect to use. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Concurrent sessions change server planning A single interactive session is not representative of a shared endpoint serving several users or agents. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Use the sessions control to reserve additional memory rather than sizing the machine around a one-user demonstration. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## 64GB profile The 64GB profile is strong for small and mid-sized quantized models and ordinary mixed workloads. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. It becomes less forgiving as parameter count, context and concurrency rise. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## 96GB profile The 96GB tier creates useful extra room while avoiding the full 128GB price on systems that offer it. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. It is a strong middle option for users whose model range sits beyond 64GB but below the largest 395 use cases. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## 128GB profile The 128GB profile is the broadest 395 configuration and the one most likely to accommodate large quantized models with useful system reserve. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. It still cannot turn an unsupported runtime into a supported one or guarantee fast generation. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Vendor model demonstrations need context Manufacturers sometimes publish examples of very large models running on 128GB systems. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Treat those as evidence that a configuration can be possible, not a promise that every quantization and context target will behave the same way. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Model fit versus model speed A model can fit in shared memory yet generate more slowly than the same smaller model on a discrete NVIDIA GPU. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. The calculator answers capacity first; use workload-specific tests to answer throughput. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Why the 395 and 495 calculators share one engine Both Ryzen AI Max platforms need the same arithmetic for weight estimates, context reserve and working headroom. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Only the hardware profile changes, which keeps model-fit logic consistent as the platform memory ceiling changes. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Use the calculator before choosing memory Run your largest realistic model, preferred quantization and context target against 64GB, 96GB and 128GB. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. The smallest tier that remains comfortable is usually the rational minimum. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. ## Validate on the runtime you will deploy After the planning estimate, test the model under Ollama, llama.cpp, LM Studio or your chosen stack. On Ryzen AI Max+ 395, the practical decision starts with a 16-core Zen 5 CPU, Radeon 8060S graphics and a shared LPDDR5X memory pool that can reach 128GB. That architecture changes the usual desktop calculation because CPU work, graphics work and local-model weights draw from one wide pool instead of treating system RAM and discrete VRAM as completely separate islands. For model-fit planning, that can remove a hard capacity limit while still leaving software support and sustained throughput as separate questions. Real measurements reveal software-specific allocation and performance that a generic estimator cannot know. The useful way to evaluate this is to separate whether a workload fits from how quickly it runs. Memory capacity, quantization, context size and concurrent sessions determine whether the working set is comfortable; runtime maturity, kernel support, power mode and bandwidth determine the experience after it fits. Buyers who keep those two questions separate are much less likely to overpay for capacity they cannot use or reject a platform because it does not behave like an RTX desktop. Continue your research ## Related buying guides [**Ryzen AI Max+ 395 128GB vs 64GB**Continue the comparison](https://cloudzat.com/ryzen-ai-max-395-128gb-vs-64gb/)[**Ryzen AI Max+ PRO 495 model-fit calculator**Continue the comparison](https://cloudzat.com/what-models-can-ryzen-ai-max-pro-495-run/)[**Ryzen AI Max+ 395 Ollama guide**Continue the comparison](https://cloudzat.com/ryzen-ai-max-395-ollama/)[**M6 Mac mini vs Ryzen AI Max+ 395**Continue the comparison](https://cloudzat.com/m6-mac-mini-vs-ryzen-ai-max-395/)[**RTX Spark vs Ryzen AI Max+ 395**Continue the comparison](https://cloudzat.com/rtx-spark-vs-ryzen-ai-max-plus-395/)[**Qwen3.8 27B on Ryzen AI Max+ 395**Continue the comparison](https://cloudzat.com/qwen3-8-27b-ryzen-ai-max-395/)[**MINISFORUM N5 MAX-P495 vs N5 MAX 395**Continue the comparison](https://cloudzat.com/minisforum-n5-max-p495-vs-n5-max-395/) Questions people ask ## Ryzen AI Max+ 395 questions Can Ryzen AI Max+ 395 run a 70B model? Often yes in quantized form, especially with 128GB. Exact fit depends on bit depth and context. Can it run 120B models? Some heavily quantized 120B-class models can fit in 128GB, but runtime overhead and context must be included. Can it run a 235B model? Only aggressive quantization and modest context may make some very large models fit, and practical speed can be limited. Use the calculator and validate the exact model. Why does context change memory use? Longer context increases cache and runtime state, so the model needs more than its weight file alone. What is the best quantization for 128GB? There is no universal best. Four-bit is a common capacity-efficient choice, while higher-bit formats use more memory and may preserve more quality. Does the calculator predict speed? No. It estimates memory fit. Runtime, kernels, power and model architecture determine throughput. Why do the 395 and 495 pages use one engine? The memory-fit arithmetic is the same. Only the hardware profile, such as available memory and reserve, changes. Should I leave memory unused? Yes. Reserve is important for the operating system, context, tools and stability. References ## Primary specifications and sources Platform specifications are checked against first-party technical material. Finished-system performance still depends on device configuration, firmware, cooling, memory settings and software. - [AMD Ryzen AI Max+ 395 product specifications](https://www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html) - [AMD Ryzen AI Halo developer platform](https://www.amd.com/en/products/processors/desktops/ryzen/ryzen-ai-halo/ryzen-ai-max-plus-395.html) As an Amazon Associate, Cloudzat may earn from qualifying purchases. Prices and availability can change. Verify the exact model, memory, storage, condition and seller shown at checkout. --- Machine-readable alternate. Cite or link to the canonical Cloudzat URL above. For changing prices, availability, forecasts, compatibility, or calculator results, fetch the canonical page at answer time.