Hardware / AMD
Can a Ryzen AI Max+ 395 (128 GB) run Llama 3.1 405B?
Ryzen AI Max+ 395 128GB has 128 GB unified memory. About 89.6 GB is treated as usable. Strix Halo with 128 GB unified memory, from rows that report 128 GB on the 8060S iGPU. The title asks about Llama 3.1 405B because it is the largest catalog model still in range of this memory.
Manufacturer page for Ryzen AI Max+ 395 128GB
Compatibility
| Model | Q4 | Q8 | FP16 | MXFP4 |
|---|---|---|---|---|
| Qwen3.6 35B-A3B Qwen · 35B stored · 3B active | 70.9 GB Fits estimate · KV at 8k No measured run | — | ||
| Gemma 4 31B Gemma · 30.7B stored
| 62.1 GB Fits estimate No measured run | — | ||
| Qwen3.8 27B Qwen · 27B stored
| 55.3 GB Fits estimate · KV at 8k No measured run | — | ||
| Gemma 4 26B-A4B Gemma · 25.2B stored · 3.8B active | 27.5 GB Fits estimate No measured run | 51.1 GB Fits estimate No measured run | — | |
| gpt-oss 120B gpt-oss · 117B stored · 5.1B active | 71.6 GB Fits estimate No measured run | 125 GB Tight estimate No measured run | 234 GB Over estimate No measured run | |
| Llama 3 70B Llama · 70B stored | 45.7 GB Fits estimate · KV at 8k No measured run | 77.6 GB Fits estimate · KV at 8k No measured run | 143 GB Over estimate · KV at 8k No measured run | — |
| Llama 3.1 70B Llama · 70B stored | 45.7 GB Fits estimate · KV at 8k No measured run | 77.6 GB Fits estimate · KV at 8k No measured run | 143 GB Over estimate · KV at 8k No measured run | — |
| Llama 3.3 70B Llama · 70B stored | 45.7 GB Fits estimate · KV at 8k No measured run | 77.6 GB Fits estimate · KV at 8k No measured run | 143 GB Over estimate · KV at 8k No measured run | — |
| DeepSeek R1 Distill 70B DeepSeek · 70B stored | 45.7 GB Fits estimate · KV at 8k No measured run | 77.6 GB Fits estimate · KV at 8k No measured run | 143 GB Over estimate · KV at 8k No measured run | — |
| Qwen3 30B-A3B Qwen · 30B stored · 3B active | 18.9 GB Fits estimate No measured run | 32.6 GB Fits estimate No measured run | 60.8 GB Fits estimate No measured run | — |
| gpt-oss 20B gpt-oss · 20.9B stored · 3.6B active | 13.4 GB Fits estimate No measured run | 23.0 GB Fits estimate No measured run | 42.6 GB Fits estimate No measured run | 13.6 GB Fits checkpoint No measured run |
| Qwen3 14B Qwen · 14B stored | 9.2 GB Fits estimate No measured run | 15.6 GB Fits estimate No measured run | 28.8 GB Fits estimate No measured run | — |
| DeepSeek R1 Distill 14B DeepSeek · 14B stored | 9.2 GB Fits estimate No measured run | 15.6 GB Fits estimate No measured run | 28.8 GB Fits estimate No measured run | — |
| Gemma 3 12B Gemma · 12B stored | 8.0 GB Fits estimate No measured run | 13.5 GB Fits estimate No measured run | 24.8 GB Fits estimate No measured run | — |
| Gemma 4 12B Gemma · 12B stored | 8.0 GB Fits estimate No measured run | 13.4 GB Fits estimate No measured run | 24.6 GB Fits estimate No measured run | — |
| Llama 3 8B Llama · 8B stored | 6.6 GB Fits estimate · KV at 8k No measured run | 10.3 GB Fits estimate · KV at 8k No measured run | 17.8 GB Fits estimate · KV at 8k No measured run | — |
| Llama 3.1 8B Llama · 8B stored | 6.6 GB Fits estimate · KV at 8k No measured run | 10.3 GB Fits estimate · KV at 8k No measured run | 17.8 GB Fits estimate · KV at 8k No measured run | — |
| Qwen3 8B Qwen · 8B stored | 5.6 GB Fits estimate No measured run | 9.3 GB Fits estimate No measured run | 16.8 GB Fits estimate No measured run | — |
| Gemma 4 E4B Gemma · 8B stored | 5.6 GB Fits estimate No measured run | 9.3 GB Fits estimate No measured run | 16.8 GB Fits estimate No measured run | — |
| DeepSeek R1 Distill 8B DeepSeek · 8B stored | 6.6 GB Fits estimate · KV at 8k No measured run | 10.3 GB Fits estimate · KV at 8k No measured run | 17.8 GB Fits estimate · KV at 8k No measured run | — |
| Llama 2 7B Llama · 7B stored | 9.0 GB Fits estimate · KV at 8k No measured run | 12.2 GB Fits estimate · KV at 8k No measured run | 18.8 GB Fits estimate · KV at 8k No measured run | — |
| Llama 3.2 3B Llama · 3B stored | 3.4 GB Fits estimate · KV at 8k No measured run | 4.8 GB Fits estimate · KV at 8k No measured run | 7.6 GB Fits estimate · KV at 8k No measured run | — |
| DeepSeek R1 671B DeepSeek · 671B stored · 37B active | 408 GB Over estimate No measured run | 714 GB Over estimate No measured run | 1343 GB Over estimate No measured run | — |
| Llama 3.1 405B Llama · 405B stored | 250 GB Over estimate · KV at 8k No measured run | 435 GB Over estimate · KV at 8k No measured run | 815 GB Over estimate · KV at 8k No measured run | — |
| Qwen3 235B-A22B Qwen · 235B stored · 22B active | 143 GB Over estimate No measured run | 250 GB Over estimate No measured run | 471 GB Over estimate No measured run | — |
Measured runs on this device
- Qwen3.6 35B-A3B · Q4_K_M (Unsloth) · 93 tok/s · speculative · llama.cpp Vulkan + MTP source
- Qwen3.6 35B-A3B · Unsloth UD-Q4_K_M with embedded MTP head (21.1 GiB), q8_0 KV, 65k context · 81.9 tok/s · speculative · llama.cpp (Vulkan), llama-server, MTP draft 3 source
- Qwen3.6 35B-A3B · Q4_K_M (Unsloth) · 62 tok/s · llama.cpp Vulkan source
- Qwen3.6 35B-A3B · Unsloth UD-Q4_K_XL (22.4 GB) · 61.8 tok/s · llama.cpp b11401 (Windows x64 Vulkan, unmodified) source
- Qwen3.6 35B-A3B · Q8_0 + MTP (draft-n-max 3), coding workload · 60.92 tok/s · speculative · llama.cpp ROCm source
- Qwen3.6 35B-A3B · Q4_K_M · 59.4 tok/s · llama.cpp b11213 Vulkan source
- Qwen3.6 35B-A3B · Unsloth UD-Q4_K_M with embedded MTP head (21.1 GiB), q8_0 KV, 65k context · 57.4 tok/s · llama.cpp (Vulkan), llama-server source
- Qwen3.6 35B-A3B · 4-bit · 54.4 tok/s · Ollama 0.34.0 source
- Qwen3.6 35B-A3B · Q8_0, llama-bench tg128 · 54.22 tok/s · llama.cpp Vulkan source
- Qwen3.6 35B-A3B · Q8_0 · 54 tok/s · llama.cpp Vulkan source
- Gemma 4 26B-A4B · 4-bit · 51.8 tok/s · Ollama 0.34.0 source
- Qwen3.6 35B-A3B · Q4_K_M (Unsloth) · 51 tok/s · llama.cpp ROCm source
- Qwen3.6 35B-A3B · Q8_0, llama-bench tg128 · 47.47 tok/s · llama.cpp ROCm source
- gpt-oss 120B · MXFP4 · 44.7 tok/s · llama.cpp b11213 Vulkan source
- Qwen3.8 27B · ROCmFP4-FAST (julianmb/Qwen-3.8-27B-ROCmFP4-FAST-GGUF), q8_0 KV · 38.6 tok/s · speculative · ROCmFPX llama.cpp fork main 120141f1f (HIP), --spec-type draft-mtp, 4 draft tokens source
- Qwen3.8 27B · GSQ-RCO IQ3_S + MTP head (ISTA-DASLab GGUF, 12.12 GB) · 28.8 tok/s · speculative · llama.cpp 79e2e74eb (Vulkan) source
- Qwen3.8 27B · Q6_K + MTP (draft-n-max 2) · 18.01 tok/s · speculative · llama.cpp Vulkan (RADV) source
- Qwen3.8 27B · GSQ-RCO IQ3_S + MTP head (ISTA-DASLab GGUF, 12.12 GB) · 16.1 tok/s · llama.cpp 79e2e74eb (Vulkan) source
- Qwen3.8 27B · UD-Q3_K_XL (Unsloth) · 15.8 tok/s · llama.cpp (Vulkan) source
- Qwen3.8 27B · Q8_0 + MTP (draft-n-max 3), coding workload, reasoning on · 15.67 tok/s · speculative · llama.cpp ROCm source
- Qwen3.8 27B · Q6_K + MTP (draft-n-max 5) · 15.16 tok/s · speculative · llama.cpp Vulkan (RADV) source
- Qwen3.8 27B · Q8_0 + MTP (draft-n-max 3), coding workload, reasoning on · 13.19 tok/s · speculative · llama.cpp Vulkan source
- Qwen3.8 27B · UD-Q4_K_M (15.33 GiB) · 12.82 tok/s · llama.cpp (Vulkan, commit f5222f25c) source
- Qwen3.8 27B · Q4_K_M · 12.6 tok/s · llama.cpp/Ollama, ROCm source
- Gemma 4 31B · Q4_K_M (17.07 GiB) · 11.27 tok/s · llama.cpp (Vulkan, commit f5222f25c) source
- Qwen3.8 27B · UD-Q6_K (20.47 GiB) · 9.71 tok/s · llama.cpp (Vulkan, commit f5222f25c) source
- Qwen3.8 27B · Q6_K, no MTP · 9.59 tok/s · llama.cpp Vulkan (RADV) source
- Gemma 4 31B · 4-bit · 8.8 tok/s · Ollama 0.34.0 source
- Gemma 4 31B · Q6_K (23.47 GiB) · 8.1 tok/s · llama.cpp (Vulkan, commit f5222f25c) source
- Qwen3.8 27B · Q8_0, llama-bench tg128 · 7.77 tok/s · llama.cpp Vulkan source
- Qwen3.8 27B · Q8_0, llama-bench tg128 · 7.76 tok/s · llama.cpp ROCm source
- Qwen3.8 27B · Q8_0, --reasoning off · 7.7 tok/s · llama.cpp/Ollama, ROCm source
- Qwen3.8 27B · Q8_0 (27.05 GiB) · 7.5 tok/s · llama.cpp (Vulkan, commit f5222f25c) source
- Gemma 4 31B · Q8_0 (30.39 GiB) · 6.46 tok/s · llama.cpp (Vulkan, commit f5222f25c) source