Just yesterday, I installed Ollama onto my 2019/20 Dell G5 laptop ( i7-9750H CPU, 16GB RAM, Nvidia GTX1660Ti 6GB GPU) running Linux Mint 22.3.I then installed Alpaca (Flatpak) and loaded the Google Gemma4 E4B model.It's small enough to completely load into GPU memory – it responds quite fast.Ok, it's not as powerful as, say Gemini Flash 3.5 running via cloud, but for a local AI it's quite usable.Today, I loaded two more models:Qwen 3.5 9B and Gemma4 12BNeither of these fitted totally within my GPU's 6GB and required the CPU to work a bit.They were thus slower (and a bit noisier).Results from all 3 were comparable, but I didn't really have a suitable test routine.Gemma4 12B is supposed to be a significantly "better" AI, but again I didn't test them thoroughly.I'm going to use Gemma4 E4B by default and reserve Gemma4 12B for more complex problems.#ai
Related
Instagram will let users endlessly swap the audio on old postsThere's a symbiotic - and sometimes frustrating - relation...
Instagram will let users endlessly swap the audio on old postsThere's a symbiotic - and sometimes frustrating - relationship between social media sites and the creators that depend...
🎮 Even sexy Skyrim mods aren't safe from AI slop anymoreBethesda's popular RPG is seeing a surge of AI-generated content...
🎮 Even sexy Skyrim mods aren't safe from AI slop anymoreBethesda's popular RPG is seeing a surge of AI-generated content, some of which has been downloaded millions of times📰 Sourc...
CodeAlmanac – Karpathy-style codebase wiki from your conversationshttps://github.com/AlmanacCode/codealmanac/Comments: h...
CodeAlmanac – Karpathy-style codebase wiki from your conversationshttps://github.com/AlmanacCode/codealmanac/Comments: https://news.ycombinator.com/item?id=48995181#HackerNews #Cod...