I'm testing local open source AI models on my simple gaming laptop using Ollama and Opencode. Only have 6 GB of VRAM with RTX 2060 so it's super slooooow. Beggars can't be choosers I guess. Either I spend the money on a better machine - signed up for email list to know when the RTX Spark chip comes out - or pay the monthly fee for something like Claude Code. Maybe the models will get better over time for those cheapskates like me? I'm using qwen3.6 but I haven't set any parameters or config to make it limit it's tokens etc yet I could probably make it a little better.