Marco Gehrke
erstmalreden
·
AI & ML interests
Local LLMs, Open-Weight Models, Agentic AI, Model Fine-Tuning, Reasoning Models, Tool Use, Multimodal AI, Model Evaluation, Quantization & Efficient Inference, AMD/Strix Halo AI
Recent Activity
Organizations
None yet
Ryzen AI Max+ 395 results: 16.68 tok/s with UD-Q5_K_XL, Vulkan and MTP 4
❤️ 3
11
#41 opened 3 days ago
by erstmalreden
Ryzen AI Max+ 395 / Windows WSL: UD-IQ2_M at 8.66–11.36 tok/s, why I rolled it back
#14 opened 3 days ago
by erstmalreden
Ryzen AI Max+ 395 results: 16.68 tok/s with UD-Q5_K_XL, Vulkan and MTP 4
❤️ 3
11
#41 opened 3 days ago
by erstmalreden