My Tiiny is set up and I'm trying to decide where to start before I fall into downloading every model in the store.
About me: I'm a developer with a small homelab: a Kubernetes cluster, a self-hosted git server, and too many dashboards. I use cloud AI coding assistants every day, and I'm keen to see how much of that work can move onto hardware I own.
What I'm planning to try:
1. One model at a time. Use it for a few days on real tasks before judging it.
2. The OpenAI-compatible API. Point my editor and a couple of homelab scripts at the Tiiny instead of the cloud.
3. A simple morning brief built from local data that never leaves the house.
Questions for people who are further along:
- Which model would you start with for everyday coding help, and which for general chat?
- Wired ethernet or Wi-Fi? Has it made a difference for you?
- What's one thing you wish you'd known in your first week?
I've already read exceltech's "Some beginner tips for Tiiny" (great post). Happy to write up what I learn as I go.
Personally, I’d start with Qwen3.6-35B-A3B for everyday coding help. It feels like a good balance of capability and responsiveness. For general chat, I’d use Qwen3.8-27B when I want something fast and conversational, and only switch to GPT-OSS-120B for tasks that genuinely benefit from the larger model. For networking, I prefer wired Ethernet through an adapter when Tiiny is staying in one place as an always-on backend. It doesn’t make inference itself faster, but model downloads, file transfers, API access, and long-running connections feel more consistent. Wi-Fi has been perfectly fine for normal chat and casual use. The main thing I wish I’d known in the first week is: don’t automatically start with the biggest model. A responsive 27B or 35B model is usually much more pleasant for daily use. I now treat Tiiny as a dedicated AI backend, start with one reliable model and workflow, and only load a larger model when the task actually needs it.