User recounts local LLM journey with llama.cpp

A Dutch user recounts running local LLMs since March 2025, starting with Mistral Nemo through SillyTavern and llama.cpp, then upgrading to dual RTX 5060 Ti 16GB cards in hopes that llama.cpp will add tensor parallel and NVFP4 support.

• France's Ecosystem Scales Agentic Computing With NVIDIA2026-10-13
• Qwen3.8 runs locally at 51 t/s on 12GB laptop GPU2026-09-30
• Four major open-weight releases in 72 hours; only NVIDIA's fits 24GB2026-10-19
• Analysis favors SK Hynix over Micron as AI memory stock2026-10-19
• Ling 3.1 Flash 560B MoE deployed across four DGX Sparks via llama.cpp2026-10-18